Factors Influencing Cryptocurrency Prices: Evidence from Bitcoin, Ethereum, Dash, Litcoin, and Monero

Journal of Economics and Financial Analysis, Vol:2, No:2 (2018) 1-27



Factors Influencing Cryptocurrency Prices: Evidence from Bitcoin, Ethereum, Dash, Litcoin, and Monero

Yhlas SOVBETOV*

Department of Economics, London School of Commerce, United Kingdom


Abstract

This paper examines factors that influence prices of most common five cryptocurrencies such Bitcoin, Ethereum, Dash, Litecoin, and Monero over 2010- 2018 using weekly data. The study employs ARDL technique and documents several findings. First, cryptomarket-related factors such as market beta, trading volume, and volatility appear to be significant determinant for all five cryptocurrencies both in short- and long-run. Second, attractiveness of cryptocurrencies also matters in terms of their price determination, but only in long-run. This indicates that formation (recognition) of the attractiveness of cryptocurrencies are subjected to time factor. In other words, it travels slowly within the market. Third, SP500 index seems to have weak positive long-run impact on Bitcoin, Ethereum, and Litcoin, while its sign turns to negative losing significance in short-run, except Bitcoin that generates an estimate of -0.20 at 10% significance level.

Lastly, error-correction models for Bitcoin, Etherem, Dash, Litcoin, and Monero show that cointegrated series cannot drift too far apart, and converge to a long- run equilibrium at a speed of 23.68%, 12.76%, 10.20%, 22.91%, and 14.27% respectively.

Keywords: Cryptocurrency; Bitcoin; Ethereum; Cointegration; ARDL Bound Test; Error Correction Model.

JEL Classification: G12, D40, C51, C59.


* Research Fellow, London School of Commerce, Chaucer House, White Hart Yard, London SE1 1NX, United Kingdom.

Email: ihlas.sovbetov @lsclondon.co.uk Tel: +44 (0)207 403 1163/ Ext.364


  1. Introduction

    Cryptocurrencies have become one of the most trending topics in recent economic and financial issues. Since Dotcom crisis, the commerce on internet (e- commerce) has been rapidly increasing and retail industries have been undergoing a revolution as internet sales are booming with more and more tech- savvy consumers go online to shop. The appetite of stock market investors for e- commerce shares seemed insatiable as investments on internet retailer were massively oversized, despite fears over the future of the Internet after the dotcom bubble burst and serious concerns about the safety of online shopping by credit cards. Until birth of first cryptocurrency -Bitcoin- in 2009, the online commerce was mainly intermediated by financial institutions serving as trusted third parties to process electronic payments. Although this system was well enough for most transactions, it was working very slowly due to controls of financial institution (problem of privacy and trust) and it was somewhat cost (transaction and commission costs).

    It triggered emerge of decentralized cryptocurrencies that bypass financial controllers, thus, transactions are very fast, smooth, and has zero cost. A cryptocurrency is defined as “a digital asset designed to work as a medium of exchange using cryptography to secure the transactions and to control the creation of additional units of the currency”. First ever use of cryptocurrency in online trade was on May 22nd 2010 by Laszlo Hanyecz, a computer programmer from Florida, for two pizzas with the amount agreed at 10,000 Bitcoins (Yermack, 2013), which would be equivalent to $155.80 million today (December 2017).

    In 2017, the popularity and use of cryptocurrencies has increased dramatically. People are "investing" vast sums of money into "assets" that have no history of producing revenue, and those assets are rising in price only because other people are also pouring money into them. Billions of dollars have been poured into more than 1,000 new digital coins issued by start-ups in 2017. These coins mimic the construction of Bitcoin, meaning they can be freely traded on digital exchanges and have no central bank standing behind them. This has raised many doubts and questions about current and future of decentralized cryptocurrencies. There are two major views. One side argues that it is a bubble with no real assets that inevitably will end with burst. The other side opines that cryptocurrency markets will become an avenue that will give millions of people an opportunity to participate in a global financial network worth tens of trillions of dollars. From young millennials in developing nations with small savings and big ambitions to mom-and-pop business owners looking to reinvest some profits in promising crypto-projects, these kinds of people will be the backbone of this industry.

    This also has increased interest of cryptocurrencies in economics and financial research sphere. Although their literature is scant, number of empirical research are growing remarkably. In this respect, we also conduct a study that examines price influences of cryptocurrencies both in short- and long-run over 2010-2018 using ARDL technique on weekly basis. As statistical data of cryptocurrency are newly established, we build the "Crypto 50" index with its total trading volume and volatility to be used in our analysis. This index is comprised of top 50 cryptocurrencies according to their market capitalisation rank. We sample most common five digital currencies such as Bitcoin, Ethereum, Litecoin, Dash, and Monero, and we scrutinize how these currencies interaction with stock market (SP500 index), gold prices, and with macroeconomic indicators (Interest rate) both in short- and long-run.


  2. Characteristics of Cryptocurrency

    A cryptocurrency is a digital or virtual currency that uses cryptography for security. A cryptocurrency is difficult to counterfeit because of this security feature. A defining feature of a cryptocurrency, and arguably its most endearing allure, is its organic nature; it is not issued by any central authority, rendering it theoretically immune to government interference or manipulation. It is designed from the ground up to take advantage of the internet and how it works. Instead of relying on traditional financial institutions that verify and guarantee your transactions, cryptocurrency transactions are verified by the user's computers logged into the currency's network. Since the currency is protected and encrypted, it becomes impossible to increase the money supply over a predefined algorithmic rate.

    One cryptocurrency, in particular, has entered the public lexicon as the go-to digital asset: Bitcoin, often is regarded as father of cryptocurrencies and all other cryptocurrencies are referred as altcoins. Since 2009, the finance world has been watching the crackerjack rise of Bitcoin with a combination of fascination and, in many cases, severe skepticism. Characteristics of Bitcoin make it fundamentally different from a fiat currency, which is backed by the full faith and credit of its government. Fiat currency issuance is a highly centralized activity supervised by a nation’s central bank. On the other hand, the value of a Bitcoin is wholly dependent on what investors are willing to pay for it at a point in time. It uses peer-to-peer blockchain network (chronologically arranged chain of blocks where each block has a list of transactions information) where all members are equal and there is no central server that tells everyone what to do (Nakamoto, 2008). This decentalisation is maintained on Satoshi Nakamoto’s (2008) idea of combining "proof of work" (PoW) with other cryptographic techniques. The PoW, mathematically, is a hash function with a large number of answer variances, the so-called "beautiful" hash is considered to be the one that is characterized by starting with 15 zeros. The hash of each block is algorithmically directly linked to the previous block. That is, if we hypothetically represent the hash function in the form,

    hash-of-current-block

    where θ is the hash of the previous block; ϕ is current difficulty level; and Z is a random key uniquely specific to the current block. This indicates that each subsequent block is inextricably linked to the previous one due to θ, and if any dishonest miner at some point decides to generate an invalid block, the other network members will not confirm it, because the hash of the previous block will not be used in it. And if spammer decides to change the hash of the previous block, then he will have to do this for the previous one as well, and so on until the genesis block (the very first block created by Satoshi Nakamoto himself). It would be incredibly time consuming to comb through the entire ledger to make sure that the person mining the most recent batch of transactions hasn't tried anything funny. This will require huge amount of work, which at the moment is almost beyond the power of one person or even a large organization. Therefore, PoW also maintains defense mechanisms for cryptcurrencies against hacking.

    However, PoW miners invest into advanced computer machines that 24/7 works (consuming energy) with the goal of validating transactions (solving hashes) and creating new blocks. Once it finds "beautiful" it declares that the block is resolved and every miner gets reward (bitcoins) proportional to their work spent on solving the hash. Therefore, cryptocurrency mining under PoW protocol is painstaking, expensive, and only sporadically rewarding. Alternatively, many altcoins started to use “proof-of-stake” (PoS) protocol which is more cost effective (cheaper) and eco-friendly (greener) comparing to PoW that requires a lots of computer energy consumption to solve mathematical algorithmic hashes. In case of PoS, miners do not need expensive computer machines, the creator of a new block is chosen in a deterministic way, depending on its wealth, also defined as stake.

    Majority of cryptocurrencies has roof limit of production. It means that supply of cryptcurrencies would decrease over time and under ceteris paribus condition should lead to higher price (inflation). However, unlike centralized fiat currencies, the cryptocurrencies are unique since their block reward schedule is public. It means that public already knows the approximate date of each decrease (or reward halving). Thus, all expectation should have been purchased by the market, and therefore shrunk in supply should not affect cryptocurrencies trading price. For instance, Bitcoin’s first block halving happened on 28th November 2012. The block reward dropped from 50 BTC per block to 25 BTC per block. The price later climbed to $260 BTC in April 2013, followed by $1,163 BTC in November 2013. It is unclear, however, whether these price rises were directly related to the block reward halving. In this research, we investigate factors that influence cryptocurrency prices both in short- and long-run.

  3. Literature Review

    The cryptocurrency market has seen an unprecedented level of interest from investors in 2016. Bitcoin, the world's largest digital currency, has risen more than 1,500 percent since the start of 2017. However, the market is significantly more complex than the public lexicon might suggest. And while there have been plenty of studies examining the future of Bitcoin and its volatility (Polasik et al. 2015; Letra, 2016; Bouoiyour and Selmi, 2016; Katsiampa, 2017; Chiu and Koeppl, 2017; Chu et al. 2017), there have been few that explore the broader cryptocurrency market and how it is evolving. Bitcoin is currently trading at around $16,000; at the beginning of the year, Bitcoin price was at $1,000, raising warnings from some analysts and prominent financial figures that it’s a bubble. The currency is extraordinarily volatile despite its recent ever-peaking performance, rising by thousands of dollars in value on one day only to fall by even more the next. Katsiampa (2017) estimates the volatility of Bitcoin through a comparison of GARCH models and finds that the AR-CGARCH model gives the most optimal fit. He underlines that the market is high speculative. Bouoiyour and Selmi (2016) study daily Bitcoin prices using an optimal-GARCH model and show that the volatility has decreasing trend comparing pre- and post-2015 data. Even tough, they still observe significant asymmetries in the Bitcoin market where the prices are driven more by negative than positive shocks. Likewise, Dyhrberg (2016) investigates the asymmetric GARCH methodology to explore the hedging capabilities of Bitcoin and he finds that it can be used as a hedging tool against stocks in the Financial Times Stock Exchange Index and against the American dollar in the short term.

    On the other hand, El Bahrawy and Alessandretti (2017) examine behaviour of entire market (1469 cryptocurrencies) between April 2013 and May 2017. They find that cryptocurrencies appear and disappear continuously and their market capitalization is increasing (super-)exponentially, several statistical properties of the market have been stable for years. Particularly, market share distribution and the turnover of crytocurrencies remain quite stable.

    There is a wide agreement on that the cryptocurrencies will not only affect the trading practices of different countries and business organizations, but they will also affect the dynamics of international relations. There are still a lot of people who are never accommodating the idea that cryptocurrencies will revolutionize how we do businesses. They can't figure out how the whole blockchain technology and other annexes work. Plus, advancements in technology are introducing digital tools that companies can use to better interact with their customers. A rising shift from traditional platforms to digital platforms has also brought about an abundant supply in data from sources like social media, mobile devices, online retail platforms, etc. Due to technology advancements in the areas of gathering, storing, and sharing data, large sets of data are easily shared among companies in every sector and country for little to no costs. The widespread accessibility of data has also brought about concerns over data privacy of individuals and their online transactions. Because every transaction or activity carried out online leaves a digital trail, individuals are opting for more anonymous ways to use the internet and conduct online transactions. The Bitcoin cryptocurrency was introduced to address the issue of privacy concern.

    Although cryptocurrencies’ decentralization, anonymity of transaction, and irreversibility of payments offer plenty advantages, Brill and Keene (2014) opine that these features also attract illegal activities (cybercriminals) such as money laundering, drug peddling, smuggling and weapons procurement. This issue has attracted the attention of powerful regulatory and other government agencies such as the Financial Crimes Enforcement Network (FinCEN), the SEC, and even the FBI and Department of Homeland Security (DHS). In March 2013, FinCEN issued rules that defined virtual currency exchanges and administrators as money service businesses, bringing them within the ambit of government regulation. In May that year, the DHS froze an account of Mt. Gox –the largest Bitcoin exchange– that was held at Wells Fargo, alleging that it broke anti-money laundering laws. And in August, New York’s Department of Financial Services issued subpoenas to 22 emerging payment companies, many of which handled Bitcoin, asking about their measures to prevent money laundering and ensure consumer protection. Plus, economist Kenneth Rogoff writes that Bitcoin will never supplant government-issued money because that “would make it extremely difficult to collect taxes or counter criminal activity.” (see Bitcoin legality in Appendix table 1A).

    To summarize, Poyser (2017) points three types of crypto price drivers organized into internal and external factors. Supply and demand of cryptcurrency is main internal factors that have direct impact on its market price. On the other hand, attractiveness (popularity), legalization (adoption), and few macro-finance factors (interest rate, stock markets, gold prices) can be regarded as external drivers (see figure 1).

    factors-influence-cryptocurrency-price

    Figure 1. Factors that Influence Cryptocurrency Prices

    In this respect, we examine short- and long-run factors that influence prices of cryptocurrencies over 2010-2018 using ARDL technique on weekly data basis. First, we build Crypto 50 index by sampling top 50 cryptocoins that have proportional contribution to market capitalization weights. Thus, we derive few cryptomarket factors such as total market capitalization, trading volume, and volatility. We use these factors as explanatory variables for cryptocoin price movements alongside with attractiveness and control variables such as stock market movements, gold prices, and interest rates. In this study, we provide evidence for significant long-run role of attractiveness of cryptocurrencies in determination of their prices. We also observe a weak form of negative impact running from stock markets (SP 500 index) to cryptocurrency market, in particular Bitcoin.

    The contents of the paper are organized as follows. Next section describes the data with descriptive analysis and explains methodological set up of examination. Then, we present our key findings including our comments and suggestions. The final section gives concluding remarks of the study.

  4. Data and Methodology

    The literature about economics of cryptocurrency is scant as the topic just recently gained focus on research fields. We contribute to this context by examining factors that influence prices of most common five cryptocurrencies over 2010-2018 with weekly data. For this examination, we define our econometric set up as following.

    ardl-long-run

    where m is optimal lag length which is determined by information criteria; Pc,t is endogenous variable in the system and it denotes price of cryptocurrency "c" in natural logarithmic form at month t. We treat all other variables in the system as exogenous variables. The Ω represents three cryptomarket variables of MARPt, MARVt, and MARSt that are Crypto 50 index price (see section 4.1), its trading volume, and its volatility at week t; and ATRc,t is attractiveness of currency "c". Plus, we also account k set of control variables of Zi such as stock market (proxied by SP500 index), exchange rates (EURO/USD), the U.S. interest rates, and world gold price.

    Data for cryptocurrencies are gathered from BitInfoCharts website1; price of SP500 index is retrieved from Yahoo Finance2, and macroeconomic data are obtained from World Bank3. The attractiveness of cryptocurrency is proxied by its Google search frequency; we derive related data from Google search trends4.


    1 BitInfoCharts - https://bitinfocharts.com/

    2 Historical Prices - https://finance.yahoo.com/world-indices

    3 https://data.worldbank.org/indicator/

    4 https://trends.google.com/trends


    1. Building Crypto 50 index

      First of all, we sample big 50 market capped cryptocurrencies (these 50 cryptocoins forms about 92% of entire cryptomarket). We derive data for market capitalization, trading volume, opening-closing prices, and high-low prices from Coinsmarketcap5. Then, we calculate weight of each cryptocoins in the index on the basis of their market capitalization. We establish Crypto 50 index (CRX50) price by summing all fifty weighted-prices as following methodology.

      crypto-50-index

      where MCi,t and Pi,t are market capitalization and price of cryptocoin i at time t respectively; and MCCRX50,t is total market capitalization of cryptocoins, that forms CRX50, at time t. We also derive daily trading volume of CRX50 index by simply summing up trading volumes of all its constituents.

      crypto-50-volume

      where VOLi,t is total trading volume of cryptocoin i at time t. Then, we derive daily volatility of our CRX50 index using formula below.

      crypto-50-volatility

      where Ph,t is the highest price of CRX50 index recorder at day t, while Pl,t is the lowest price of CRX50 index recorded at day t. The high-low price of CRX50 index is derived by methodology explained in equation 2.

    2. Brief Overview of Cryptocurrency Market

      We briefly summarize economics of cryptocurrency market by outlining key statistics. Coin Dance6 regularly announces up-to-date and historical report statistics of cryptocurrency markets. According their most recent report, 34.4% of total market share belongs to Bitcoin, while 19.23%, 10.74%, and 1.97% shares are attributable to Ethereum, Ripple, and Litcoin respectively. Moreover, their report shows that 96.57% of cryptocurrency market involvers are males, while only 3.43% are females. The age distribution refers to ability of the cryptocurrency market to attract wide range of people from very young to very old. The report gives statistics for only 18+ ages where 8.36% of the market involvers are aged 18- 24, while 45.71% and 30.62% are attributable to people aged 25-34 and 35-44 respectively. Interestingly, the share of elderly people (45+) is about 16% which provides evidence for that cryptocurrency market attracts from youngest to elderly people into financial activities. The cryptocurrency interest and affinity statistics also show that people of cryptocurrency community are mainly engaged with financial activities, pursing investment opportunities.


      5 https://coinmarketcap.com/

      6 https://coin.dance


      cryptocurrency-market-share-engagement-demographics

      Figure 2. Cryptocurrency Market Share and Engagement Demographics


      coin-dance-demographics-interests


      coin-dance-demographics-affinities

      Figure 3. Bitcoin Community Interest and Affinity


      On the other hand, Google search frequency for "Bitcoin" (or "BTC") and "Blockchain" terms also shows fairly significant correlation with Bitcoin and Altcoins prices respectively (see figure 4). This seems to be a significant explanatory factor of cryptocurrency prices; therefore, we use this indicator to proxy attractiveness of cryptocurrency in this study.


      coin-dance-market-cap-historical


      coin-dance-bitcoin-trend


      coin-dance-blockchain-trend

      Figure 4. Bitcoin-Altcoin Market Cap vs Bitcoin-Blockchain Google Search Trends

      To summarize, we briefly outline some cryptocurrency-specific figures in table 1, plus, we describes our data with brief abbreviations and statistics in table 2.

      Table 1. Overview of Cryptocurrency Market


      BitcoinEthereumDashLitecoinMonero
      Total16.7 million BTC96.8 million ETH7.8 million DASH54.7 million LTC15.6 million XMR
      Price (USD)$ 14,729.86$ 1,082.47$ 1,067.01$ 248.93$ 389.18
      Market Cap. (USD)$248 billion$106 billion$8 billion$13 billion$6 billion
      Transactions / hour13,60949,9006146,036220
      Sent / hour121,019 BTC
      ($1.8 billion USD)
      558,178 ETH
      ($612 million USD)
      9,276 DASH
      ($10 million USD)
      559,875 LTC
      ($140 million USD)
      683,751 XRP
      ($1.63 million USD)
      Avg. Transaction Value8.89 BTC
      ($131,519 USD)
      11.19 ETH
      ($12,273 USD)
      15.10 DASH
      ($16,146 USD)
      92.76 LTC
      ($23,174 USD)
      75.28 XMR
      ($29,297 USD)
      Median Transaction
      Value
      0.366 BTC
      ($5,416.88 USD)
      0.197 ETH
      ($216.3 USD)
      0.605 DASH
      ($647.12 USD)
      10.83 LTC
      ($2,706.15 USD)
      9.35 XMR
      ($3638.33 USD)
      Block Time9m 17s15.8s2m 37s2m 33s2m 0s
      Blocks Count503,1894,875,208800,8141,346,8321,487116
      Blocks last 24h1545460548562713
      Blocks / hour6228232330
      Reward Per Block12.50 BTC
      ($246,517 USD)
      3 ETH
      ($4,744 USD)
      3.60 DASH
      ($3,867USD)
      25 LTC
      ($6,321 USD)
      5.43 XMR
      ($2,186 USD)
      Difficulty1.93114*10122.003*101570.58*1063.7*10675.8 * 109
      Hashrate (Hash/second)15.58 * 1018169.10 *10121.93 *101599.98 *10126.26 * 108
      Mining Profitability/Day2.4364 USD0.1531 USD0.5493 USD0.0355 USD2.4905 USD
      Wealth Distribution
      Top 10 addesses
      Top 100 addesses
      Top 1,000 addesses
      Top 10,000 addesses
      10 - 5.25%
      100 - 17.89%
      1000 - 34.25%
      10000 - 55.66%
      10 - 10.82%
      100 - 33.90%
      1000 - 53.75%
      10000 - 69.61%
      10 - 6.32%
      100 - 15.64%
      1000 - 28.53%
      10000 - 92.37%
      10 - 14.44%
      100 - 48.61%
      1000 - 65.91%
      10000 - 79.96%
      10 - 18.03%
      100 - 51.17%
      1000 - 71.85%
      10000 - 84.29%
      100 Largest
      Transactions in
      Last 24h
      713,840 BTC
      ($10.56 billion)
      24.58% Total
      1,055,897 ETH
      ($1.1 billion)
      7.88% Total
      128,562 DASH
      ($137 million)
      57.75% Total
      2,806,164 LTC
      ($701 million)
      20.88% Total
      2,042,328 XMR
      ($794 million)
      13.25% Total
      First Block (Genesis)2009-01-092015-07-302014-01-192011-10-082014-04-18
      Blockchain Size178.49 GB293.11 GB5.04 GB12.69 GB39.22 GB

      Table 2. Descriptive Statistics

         SeriesAbbrMeanMedianMax.Min.Std.Dev.SkewnessKurtosisN
      PRICEBitcoin priceBITPP4.435.489.72-2.812.86-0.762.86390
      Ethereum priceETHP2.912.486.87-0.802.080.152.03126
      Dash priceDASP2.281.857.35-0.531.841.013.01203
      Litcoin priceLITP1.831.365.750.151.161.243.85245
      Monero priceMONP0.63-0.156.15-3.992.410.432.15193
      EURO/USD PriceEURP1.251.281.481.040.12-0.131.65390
      Gold PriceGOLP7.217.177.516.960.140.572.27390
      SP500 PriceSPP7.467.537.916.960.24-0.231.81390
      MARKET CAPBitcoin mar.capBITM12.343.47297.530.0034.995.6739.72390
      Ethereum mar.capETHM9.800.9985.700.0317.022.288.65126
      Dash mar.capDASM0.540.039.550.001.464.3123.70203
      Litcoin mar.capLITM0.780.1819.010.042.236.0343.28245
      Monero mar.capMONM2.190.2495.460.037.748.7497.55193
      TRADING VOLUMEBitcoin volumeBITV4.504.0410.012.251.751.153.60210
      Ethereum volumeETHV3.453.038.54-1.612.600.072.18126
      Dash volumeDASV-0.42-1.256.71-4.652.620.852.59203
      Litcoin volumeLITV2.001.268.13-0.371.961.313.63210
      Monero volumeMONV0.63-0.156.15-4.002.410.432.15193
      MARKETCRX50 priceMARP4.315.379.23-2.812.75-0.862.96390
      CRX50 mar.capMARM-0.391.376.45-11.334.10-1.013.34390
      CRX50 volumeMARV-2.04-2.543.99-4.321.971.203.51210
      CRX50 volatilityMARS-3.11-3.210.04-9.871.07-0.276.74390
      ATTRACTIVENESSBitcoin trendBITA0.780.782.000.300.341.324.92390
      Ethereum trendETHA0.140.002.000.000.403.239.85126
      Dash trendDASA0.150.002.000.000.303.4210.87203
      Litcoin trendLITA0.100.002.000.000.323.4912.87245
      Monero trendMONA0.310.302.000.000.214.7533.38193

    3. Model Specification

      Prior to cointegration analysis, we should make sure that variables are integrated at same degree. We examine characteristics of all series by employing Augmented Dickey- Fuller (ADF) unit root test as following.

      augmented-dickey-fuller-adf

      where ΔΩt is the first difference of a variable Ω; Τ is a trend, and θ1 is its multiplier; k is an optimal lag length; and εt is a White Noise residual term. Here, ADF hypothesizes H0 (ρ=0) against alternative (ρ<0), and rejection of the null confirms stationarity of Ω.

      We display results of ADF test in table 3 where we find majority of series are non-stationary at level. But they can be converted to a stationary through first differencing methodology. Thus, we conclude that all series are I(1) variables, except volatility (sigma) and few attractiveness series that are seem to be I(0). Finding series are not integrated in same degree, we decide to use Autoregressive Distributed Lag (ARDL) cointegration framework which is also known as Bound testing approach. This technique is applicable for series with mixture of I(0) and I(1) variables, but none of them should be I(2).

      Table 3. Output of ADF Analysis


      Series

      Level

      First Difference

      Prob.

      Lag

      Max Lag

      N

      Prob.

      Lag

      Max Lag

      N

      PRICE

      BITPP

      0.3360

      1

      16

      388

      0.0000

      0

      16

      388

      ETHP

      0.5106

      2

      12

      123

      0.0000

      1

      12

      123

      DASP

      0.9654

      0

      14

      202

      0.0000

      0

      14

      201

      LITP

      0.9873

      0

      15

      244

      0.0000

      0

      15

      243

      MONP

      0.2991

      2

      14

      190

      0.0000

      1

      14

      190

      EURP

      0.4698

      0

      16

      389

      0.0000

      0

      16

      388

      GOLP

      0.2050

      0

      16

      389

      0.0000

      0

      16

      388

      SPP

      0.2148

      1

      16

      388

      0.0000

      0

      16

      388

      MARKET

      CAP

      BITM

      0.1352

      3

      16

      386

      0.0238

      2

      16

      386

      ETHM

      0.1927

      3

      12

      122

      0.0003

      1

      12

      123

      DASM

      0.9847

      0

      14

      202

      0.0375

      1

      14

      200

      LITM

      0.9272

      2

      15

      242

      0.0458

      2

      15

      241

      MONM

      0.8451

      0

      14

      190

      0.0174

      0

      14

      189

      TRADING

      VOLUME

      BITV

      0.1214

      1

      14

      208

      0.0000

      3

      14

      205

      ETHV

      0.0231

      0

      12

      125

      0.0000

      0

      12

      124

      DASV

      0.6298

      2

      14

      200

      0.0000

      1

      14

      200

      LITV

      0.6355

      2

      14

      207

      0.0000

      1

      14

      207

      MONV

      0.4942

      0

      14

      190

      0.0000

      1

      14

      188

      MARKET

      MARP

      0.2931

      1

      16

      388

      0.0000

      0

      16

      388

      MARM

      0.1334

      3

      16

      386

      0.0000

      2

      16

      386

      MARV

      0.7168

      2

      14

      207

      0.0000

      3

      14

      205

      MARS

      0.0000

      1

      16

      388

      0.0000

      5

      16

      383


      ATTRACTIVENESS

      BITA

      0.0732

      2

      16

      388

      0.0051

      1

      16

      388

      ETHA

      0.0978

      1

      12

      124

      0.0308

      0

      12

      124

      DASA

      0.1433

      2

      14

      200

      0.0165

      0

      14

      201

      LITA

      0.1340

      2

      15

      242

      0.0334

      1

      15

      242

      MONA

      0.1143

      0

      15

      190

      0.0291

      0

      15

      189

      Notes: The numbers are F-statistics derived with ADF unit root test using levels and first differences. The lag length criterion is set as Schwarz (1978) Information Criterion (SIC) with automatic maximum 25 lags. The tested model includes individual effects and individual linear trends.


      The bound testing methodology, is pioneered by Pesaran et al. (2001), tests potential cointegration of I(0) and I(1) variables in long-run. The technique also provides some evidence for series short-run and error-correction dynamics. Thus, we recall equation 1, and adjust it in accordance to the ARDL approach following Pesaran et al.(2001) as below.

      ardl-model-equation

      where X stands for four cryptocurrency-related variables of MARP, MARV, MARS, and ATR; and Z stands for four control variables of SPP, EURP, GOLP, and INT. The lag of dependent variable starts from 1 to its optimal lag length (m). However, the independent variables begin from lag zero and continue up to their optimal, i.e. n1-n8, which are determined by Schwartz (1978) Information Criterion (SIC).

      Thus, the null hypothesis of φ1= φ2j3j=0 is tested with Wald analysis where rejection of H0, under Pesaran et al (2001) lower and upper bound critical values, indicates existence of long-run cointegration between series only if the residual of equation 1.0 model (εt) is stationary. In case of justification of these requirements the Restricted Error Correction Model (RECM) can be formulated as below.

      ardl-error-correction-model

      where ECTc is White-noise stationary residual of long-run equation 2 (εt), and λc is its multiplier that is expected to be statistically significant in the range of -1 and 0 for robustness of RECM model (3). In case, λc is estimated positive, then the model is suffering of serially correlated residual terms (autocorrelation problem). And if λc is estimated negative but greater than 1 (in absolute terms), then the model is instable comprising of structural breaks that are needed to be controlled (Sovbetov and Saka, 2018).

  5. Findings

    Firstly, we ensure that residual of equation 1 model (εt) is stationary at 1% significance level. Next, we employ an ARDL optimal lag selection test in Eviews 9.0 software for equation 3 model, setting maximum lag length as 4 under SIC. As a result, the test finds ARDL(3,1,1,0,0,0,0,0,0,0) specification as most appropriate for BITP model where SIC value (10.8776) is the minimum (see Figure 5, Panel A). In other words, the SIC suggests that our ARDL model should include only three lags of dependent variables (Pc), one lag of MARP and MARV. For robustness of this ARDL model, we examine its residual under Serial Correlation LM and Heteroskedasticity test. We find that residuals do not comprise these two problems as probabilities of Chi-Square statistics of Breusch-Godfrey Serial Correlation LM test and Breusch-Pagan-Godfrey Heteroskedasticity test are greater than 11% and 12% respectively. However, the percentages are quite close to 10% significance level, therefore, we use HAC-robust standard errors in ARDL model. Further, we examine the stability of the model by employing CUSUM test that checks changes in cumulative sum of recursive residuals over time. In panel B of the figure 5, we demonstrate graphical outcome of CUSUM test where clearly seen that CUSUM (blue) line does not exceed ±5% significance (two red) lines, indicating stability of our model over the analysis period.


    graph_ardl_sel_bitp graph_cusum_bitp

    Panel A. ARDL Lag Selection with SIC     Panel B. CUSUM Test Result for ARDL Model

    Figure 5. ARDL Lag Selection for BITP model with SIC & Stability of Selected ARDL Model


    graph_ardl_sel_ethp graph_cusum_eth

    Panel A. ARDL Lag Selection with SIC     Panel B. CUSUM Test Result for ARDL Model

    Figure 6. ARDL Lag Selection for ETHP model with SIC & Stability of Selected ARDL Model


    graph_arld_sel_dasp graph_cusum_dasp

    Panel A. ARDL Lag Selection with SIC     Panel B. CUSUM Test Result for ARDL Model

    Figure 7. ARDL Lag Selection for DASP model with SIC & Stability of Selected ARDL Model


    graph_ardl_sel_litp graph_cusum_litp

    Panel A. ARDL Lag Selection with SIC     Panel B. CUSUM Test Result for ARDL Model

    Figure 8. ARDL Lag Selection for LITP model with SIC & Stability of Selected ARDL Model

    graph_ardl_sel_monp graph_cusum_monp

    Panel A. ARDL Lag Selection with SIC     Panel B. CUSUM Test Result for ARDL Model

    Figure 9. ARDL Lag Selection for MONP model with SIC & Stability of Selected ARDL Model

    Similarly, on basis of minimum SIC value, we select the most appropriate ARDL models for ETHP (see figure 6), DASP (see figure 7), LITP (see figure 8), and MONP (see figure 9).

    After specifying most appropriate models for our five cryptocurrency, we run equation 1 for each models separately. Table 4 reports outcome of these analyses where we observe plausible results. These models comprise both short- and long- run dynamics. In all cases Pt-1 derives statistically significant estimates at 1% level. We can derive long-run multipliers of related variables by using formula of i1 where i has an array of {21, 22, 23, 24, 31, 32, 33, 34} that corresponds to {MARP, MARV, MARS, ATR, EURP, GOLP, SPP, INT}. However, we first need to carry out Wald test hypothesizing H0: φi=0, to examine whether these series (cryptocurrency market variables and control variables) have statistically significant long-run interactions with Pt (cryptocurrency BITP, ETHP, DASP, LITP, and MONP). We report results of Walt test in table 5 for each cryptocurrency where critical values for lower (I(0)) and upper (I(1)) bounds of each case also are given. Case I should be used for models that include neither intercept nor trends of any kind. Case II should be used for models that comprise only restricted intercept without any trends. In other words, the intercept is allowed only in long- run (Unrestricted ECM) model, but not short-run (Restricted ECM) model. On the other hand, Case III allows unrestricted intercept, but no any trends. In this case, short-run model has an intercept and no trends. Case IV should be used for models that comprise both unrestricted intercept and restricted trend. So that short-run specification of these models includes intercept, while long-run specification comprises trend factor.

    Therefore, specifying short- and long-run ARDL models is important. In this respect, we follow simple logic, we look if inclusion of intercept and trend add statistically significance into models or not. Once, we find they are statistically significant; we left them in the model. Following this methodology, we observe that trend factor appears insignificant in all cases (BITP, ETHP, DASP, LITP, and MONP). Thus, we disregard it in all cases. On the other hand, intercepts in BITP, LITP, and MONP models appear significant only in long-run, but not in short-run. Therefore, we choose case II specification for these three models. Intercept appears statistically insignificant only in DASP model, consequently removing it from the model we end up with Case I specification. Lastly, we observe that intercept in unrestricted in short-run model of ETHP as it derived statistically significance at 1% level. Thus, we choose case III specification for ETHP model.

    Table 4. Results of ARDL Models

    Variable

    ΔBITP

    ΔETHP

    ΔDASP

    ΔLITP

    ΔMONP

    ΔPt-1

    0.4208*** (0.0851)

    -

    0.2362** (0.1165)

    0.2188*** (0.0799)

    0.9310*** (0.1865)

    ΔPt-2

    0.0260** (0.0131)

    -

    0.3307** (0.1422)

    0.2297** (0.1187)

    -

    ΔMARPt

    0.6966*** (0.0715)

    0.2111*** (0.0313)

    0.7837*** (0.0117)

    0.7573*** (0.0084)

    0.0484*** (0.0071)

    ΔMARVt

    0.1312*** (0.0328)

    -

    -

    -

    0.0061*** (0.0015)

    MARPt-1 (φ21)

    0.0676*** (0.0209)

    0.0839*** (0.0320)

    0.0944** (0.0375)

    0.1419** (0.0618)

    0.1101** (0.0562)

    MARVt-1 (φ22)

    0.0121** (0.0055)

    0.0287** (0.0124)

    0.0107*** (0.0041)

    0.0265* (0.0154)

    0.0143

    (0.0187)

    MARSt-1 (φ23)

    -0.0128

    (0.0139)

    -0.0328

    (0.0364)

    -0.0205

    (0.0214)

    -0.0090

    (0.0162)

    -0.0028

    (0.0032)

    ATRt-1 (φ24)

    0.1085*** (0.0311)

    0.0521*** (0.0162)

    0.0328

    (0.0493)

    0.0288*** (0.0101)

    0.0205

    (0.0171)

    EURPt-1 (φ31)

    0.0628

    (0.0428)

    0.0376

    (0.0909)

    0.0126

    (0.0344)

    0.0164

    (0.0380)

    0.0133

    (0.0349)

    GOLPt-1 (φ32)

    -0.0333

    (0.0439)

    0.0131

    (0.0374)

    -0.0023

    (0.0682)

    0.0109

    (0.0597)

    0.0092

    (0.0401)

    SPPt-1 (φ33)

    0.0696* (0.0368)

    0.0467* (0.0251)

    0.0349

    (0.0316)

    0.0165* (0.0103)

    0.0122

    (0.0169)

    INTt-1 (φ34)

    -0.0317

    (0.0535)

    -0.0128

    (0.0702)

    -0.0049

    (0.0172)

    -0.0189

    (0.0328)

    -0.0001

    (0.0007)

    Pt-1 (φ1)

    -0.0851*** (0.0299)

    -0.2181*** (0.0529)

    -0.8801*** (0.1179)

    -0.4127*** (0.0780)

    -0.4331*** (0.1091)

    intercept

    -0.0391*** (0.0118)

    -0.0597*** (0.0276)

    -

    0.0325*** (0.0035)

    0.0404*** (0.0060)

    R-squared

    0.7122

    0.5929

    0.4703

    0.5791

    0.5454

    DurbinWatson

    2.0109

    1.9282

    2.0629

    1.9929

    1.9489

    BG LM Test

    0.1095

    0.1413

    0.0782

    0.1276

    0.0391

    BPG Test

    0.1216

    0.0738

    0.0655

    0.0588

    0.1302

    Notes: Numbers in the table are estimations derived by ARDL technique with maximum 4 lags allowance. We use SIC in lag length selection. The standard errors are in HAC-robust characteristics with Bartlett kernel and Newey-West (1987) fixed bandwidth 5.

    Table 5 presents Wald test of our five models comparatively with Pesaran et al (2001) critical values of lower and upper bounds for k=8. This k indicates number of original regressors in the model, except the dependent variable. The table shows that computed F-statistics exceeds critical bound values even at 1% significance level in all cases, signifying strong long-run cointegrations among mentioned series.

    Table 5. F-test with Bound Critical Values

    10% level

    5% level

    1% level


    Bounds

    Lower Bound
    [I(0)]

    Upper Bound
    [I(1)]

    Lower Bound
    [I(0)]

    Upper Bound
    [I(1)]

    Lower Bound
    [I(0)]

    Upper Bound
    [I(1)]

    Wald Test

    Case I

    1.66

    2.79

    1.91

    3.11

    2.45

    3.79

    DASP (5.04***)


    Case II


    1.85


    2.85


    2.11


    3.15


    2.62


    3.77

    BITP (8.96***)
    LITP (6.40***)
    MONP (5.28***)

    Case III

    1.95

    3.06

    2.22

    3.39

    2.79

    4.1

    ETHP (7.53***)

    Case IV

    2.13

    3.09

    2.38

    3.41

    2.93

    4.06


    Notes: The critical values for each case are retrieved from Pesaran et al (2001) Table CI. The k indicates the number original regressors in the model. Therefore, it is 8 for in equation 1 model (disregarding dependent variable Pt-1). Wald test hypothesizes null H0: φi=0. The significance levels respectively as *:10%, **:5%, and ***:1%.

    Now, we can estimate approximate magnitude of cointegrations (long-run relationships) by calculating negative ratio of coefficients of independent variables to dependent one (i1). Table 6 shows these long-run multipliers automatically derived by Eviews software with their HAC-robust standard errors in parenthesis. According to results, we document that long-run market beta (coefficient of MARP) is statistically significant at 1% level in Bitcoin and Ethereum models where its multiplier is 0.79 and 0.38 respectively. Whereas, Dash, Litcoin, and Monero models predict it as 0.11, 0.34, and 0.25 respectively at 5% significance level. We believe that these results emerge as Bitcoin and Ethereum comprise the largest market share of entire cryptocurrency market, and their beta coefficient shows higher responsiveness to the market in long-run. In other words, 1 unit increase in MARP leads Bitcoin and Ethereum to increase by 0.79 and 0.38 units respectively.

    On the other hand, trading volume appears to have significant long-run impact on Bitcoin at 1% significance level and on Ethereum, Litcoin, and Monero at 10% significance level. In case of Dash model, it appears statistically insignificant. The result indicates that a unit increase in weekly trading volume causes 0.14, 0.13, 0.06, and 0.03 increases in Bitcoin, Ethereum, Litcoin, and Monero cryptocurrencies in long-run. Sigma, proxied by volatility of the cryptocurrency market, emerges statistically significant long-run impact on all cryptocurrencies. The sign of impact is negative, which indicates a unit increase in volatility of the market causes Bitcoin to drop by 0.15 units, Ethereum by 0.15 units, Dash by 0.02, Litcoin by 0.02 units, and Monero by 0.01 units in long-run. In addition, we observe that attractiveness (proxied by Google search term frequency) also derives significant coefficients for Bitcoin and Ethereum at 1% significance level and for Litcoin and Monero at 10% significance level. It indicates that 1 unit increase in Google trend popularity of Bitcoin, Ethereum, Litcoin, and Monero leads 1.27, 0.24, 0.07, and 0.05 units increases in their long-run prices respectively. Google search frequency appears to be insignificant factor for Dash.

    Coming to macroeconomic control variables, we observe that majority of them seem to be statistically insignificant factor in explaining price movements in cryptocurrencies. Only, SP500 index derives weak form of significant coefficient (10% level) in Bitcoin, Ethereum, and Litcoin models. The positive sign of SPP indicates that a unit increase leads 0.81, 0.21, 0.04 raise in Bitcoin, Ethereum, and Litcoin prices respectively in long-run. The logic behind this relationship appears ambiguous. Normally, one could expect a stronger USD against other fiat currencies (including cryptocurrencies) when SPP increases.

    Table 6. Long-run estimates of Cryptocurrency ARDL models


    BITP

    ETHP

    DASP

    LITP

    MONP

    MARP

    0.7944*** (0.0599)

    0.3847*** (0.0654)

    0.1073** (0.0519)

    0.3438** (0.1649)

    0.2542** (0.1202)

    MARV

    0.1425*** (0.0349)

    0.1316* (0.0713)

    0.0122

    (0.0177)

    0.0642* (0.0362)

    0.0330* (0.0187)

    MARS

    -0.1511*** (0.0526)

    -0.1504*** (0.0564)

    -0.0233* (0.0125)

    -0.0218** (0.0102)

    -0.0065** (0.0032)

    ATR

    1.2750*** (0.1511)

    0.2389*** (0.0721)

    0.0373

    (0.0328)

    0.0698** (0.0334)

    0.0473** (0.0197)

    EURP

    0.7381*** (0.2446)

    0.1724

    (0.1149)

    0.0143

    (0.0385)

    0.0397

    0.0502)

    0.0307

    (0.0349)

    GOLP

    -0.3913

    (0.2788)

    0.0601

    (0.0795)

    -0.0026

    (0.0681)

    -0.0264

    (0.0681)

    -0.0212

    (0.0401)

    SPP

    0.8179* (0.4250)

    0.2141* (0.1133)

    0.0397

    (0.0324)

    0.0400* (0.0213)

    0.0282

    (0.0184)

    INT

    -0.3725

    (0.2357)

    -0.0587

    (0.0683)

    -0.0056

    (0.0243)

    -0.0458

    (0.0464)

    -0.0002

    (0.0171)

    C

    0.4596*** (0.0827)

    -

    -

    0.3668*** (0.0739)

    0.4284*** (0.0817)

    Notes: Estimates are derived by long-run unrestricted ARDL technique with HAC-robust standard errors in parenthesis. The significance levels are: 10% (*), 5% (**), and1% (***).

    Further, we estimate short-run error-correction equation (4) and we report outcome of this analysis in table 7. Apparently, all ECM models generate consistent coefficients. The coefficient of MKT implies that a unit increase in cryptocurrency market return causes Bitcoin, Ethereum, Dash, Litcoin, and Monero to increase by 0.85, 0.39, 0.04, 0.12, and 0.09 units respectively in short-run. Notice that short-run coefficients of Bitcoin and Ethereum are higher comparing to their long-run coefficients, indicating that their responses are more sensitive in the short-run. Likewise, a unit increase in cryptcurrency market trading volume leads mentioned cryptocurrencies to increase by 0.03, 0.01, 0.007, 0.005, and 0.004 units respectively at 1%-5% significance level. These short-run coefficients seem to be lesser than their long-run magnitudes at table 6, indicating that responses of the cryptocurrencies to the fluctuations in market trading volume are higher in long-run.

    Table 7. Short-run estimates of ARDL Error-Correction Cryptocurrency models


    BITP

    ETHP

    DASP

    LITP

    MONP

    ΔPt-1

    0.2087*** (0.0429)

    -

    0.1585*** (0.0545)

    0.3973*** (0.0651)

    0.5271*** (0.1163)

    ΔPt-2

    0.1251*** (0.0115)

    -

    0.1162*** (0.0322)

    0.1862*** (0.0642)

    -

    ΔMARP

    0.8485*** (0.0983)

    0.3914*** (0.0457)

    0.0414** (0.0187)

    0.1195** (0.0531)

    0.0874** (0.0428)

    ΔMARV

    0.0315*** (0.0022)

    0.0113*** (0.0011)

    0.0075** (0.0034)

    0.0053** (0.0027)

    0.0041** (0.0020)

    ΔMARS

    -0.3896** (0.1902)

    -0.2409* (0.1453)

    -0.1695* (0.0913)

    -0.1991** (0.0977)

    -0.1347* (0.0735)

    ΔATR

    0.1372* (0.0817)

    0.0538

    (0.0644)

    0.0270

    (0.0451)

    0.0257

    (0.0211)

    0.0288

    (0.0436)

    ΔEURP

    0.0723

    (0.1131)

    0.0473

    (0.0899)

    0.0148

    (0.0595)

    0.0245

    (0.0395)

    0.0280

    (0.0409)

    ΔGOLP

    0.1399

    (0.1664)

    -0.0023

    (0.0962)

    0.0768

    (0.1452)

    0.0462

    (0.0565)

    0.0301

    (0.0403)

    ΔSPP

    -0.2020* (0.1328)

    -0.0683

    (0.0676)

    0.0410

    (0.0718)

    -0.0360

    (0.0328)

    -0.0203

    (0.0302)

    ΔINT

    -0.0597

    (0.1106)

    -0.0234

    (0.0844)

    -0.0022

    (0.0087)

    -0.0175

    (0.0314)

    -0.0019

    (0.0058)

    C

    -

    1.1541***
    (0.2118)

    -

    -

    -

    ECTt-1

    -0.2368***

    (0.0361)

    -0.1276***

    (0.0303)

    -0.1020***

    (0.0212)

    -0.2291***

    (0.0472)

    -0.1427***

    (0.0335)

    Notes: The estimates are derived by short-run RECM ARDL model with HAC-robust standard errors in parenthesis. ECTt-1 shows speed of adjustment towards long-run equilibrium. The significance levels are: 10% (*), 5% (**), and1% (***). For data description and abbreviations see table 2.

    Market volatility also appears statistically significant and negatively signed as it was in long-run. Moreover, its short-run magnitudes are several-fold comparing to their long-run magnitudes, signalizing that the cryptocurrencies show more severe reaction to the market's volatility in short-run.

    Interestingly, attractiveness factor derives insignificant estimates in almost all models, except Bitcoin that predicts 0.14 coefficient at 10% significance level. This indicates that impact of attractiveness on cryptocurrency is subjected to time factor, indicating that it has latent characteristics and its formation (may be also recognition by the market) requires a time.

    On the other hand, macroeconomic control variables appear insignificant in all short-run models, except Bitcoin model that predicts estimate of SPP factor as - 0.2020 at 10% significance level. This indicates that a unit increase in SP500 index causes Bitcoin prices to decrease by 0.20 units in short-run. An inverse relationship is documented in long-run with absolute magnitude of several-fold. This, indeed, confirms negative correlation between Bitcoin prices and SP500 index in short- run.

    Lastly, error correction terms (ECT) in all models appear statistically significant at 1% level with negative sign complying with the ECM theory. Bitcoin model seems to be correcting 23.68% of its previous period disequilibrium in the way converging its long-run level. This adjustment speed in Etherem, Dash, Litcoin, and Monero is 12.76%, 10.20%, 22.91%, and 14.27% respectively.

  6. Concluding remarks

    This paper examines factors that influence prices of most common five cryptocurrencies such Bitcoin, Ethereum, Dash, Litecoin, and Monero over 2010- 2018 using weekly data and documents several results. First, using differencing methodology to stationarize series wipes out potential long-run interactions between series; therefore, we use Autoregressive Distributed Lag (ARDL) technique in order to account both short- and long-run dynamics of cryptocurrency prices as our data sample is comprised of mixture of I(0) and I(1) variables. Unrestricted long-run ARDL and restricted short-run error-correction analyses find statistically significant impact running from cryptomarket factors such as total market prices, trading volume, and volatility on to five cryptocurrencies in long- and short-run respectively.

    The cryptomarket beta derives a long-run multiplier of 0.79 on Bitcoin and 0.38 on Ethereum at 1% significance level, while it generates 0.11, 0.34, and 0.25 long-run impacts on Dash, Litcoin, and Monero at 5% significance level. This indicates that Bitcoin and Ethereum have higher responsiveness to the market in long-run. In case of short-run, a unit increase in cryptocurrency market return causes Bitcoin, Ethereum, Dash, Litcoin, and Monero to increase by 0.85, 0.39, 0.04, 0.12, and 0.09 units respectively in short-run. As short-run multiplier of Bitcoin and Ethereum are greater than their long-run coefficients, we conclude that these responses of these two cryptocurrencies are more sensitive in short-run.

    Trading volume appears to have significant long-run impact on Bitcoin at 1% significance level and on Ethereum, Litcoin, and Monero at 10% significance level, indicating a unit increase in weekly trading volume causes 0.14, 0.13, 0.06, and 0.03 raises in Bitcoin, Ethereum, Litcoin, and Monero cryptocurrencies in long-run. In case of short-run dynamics, all five cryptocurrencies earn statistically significant estimates. However, these estimates seem to be lesser than their long-run magnitudes, indicating that responses of the cryptocurrencies to the fluctuations in market trading volume are higher in long-run.

    Likewise, volatility of the cryptocurrency market appears to be statistically significant determinant both in long- and short-runs for all cryptocurrencies. The sign of impact is negative, which indicates a unit increase in volatility of the market causes Bitcoin to drop by 0.15 units, Ethereum by 0.15 units, Dash by 0.02, Litcoin by 0.02 units, and Monero by 0.01 units in long-run. In case of short-run, these impacts seem to be several-fold, indicating that the cryptocurrencies show more severe reaction to the market's volatility in short-run.

    Attractiveness of cryptocurrencies also matters for all except Dash, but only in long-run. It derives significant coefficients for Bitcoin and Ethereum at 1% significance level and for Litcoin and Monero at 10% significance level, indicating that 1 unit increase in attractiveness of Bitcoin, Ethereum, Litcoin, and Monero leads 1.27, 0.24, 0.07, and 0.05 units increases in their long-run prices respectively. In case of short-run analysis, attractiveness factor derives insignificant estimates for almost all models, except Bitcoin that earns an estimate of 0.14 at 10% significance level. This indicates that formation and recognition of the attractiveness of cryptocurrencies are subjected to time factor. In other words, they travel slowly within the market.

    In case of control variables, SP500 index derives weak form of positive significant coefficient (10% level) in Bitcoin, Ethereum, and Litcoin models. Although the logic behind these positive long-run relationships appears ambiguous, they totally disappear in short-run, while only Bitcoin model predicts a negative estimate that is statistically significant 10% significance. This confirms that one could expect a stronger USD against other fiat currencies (including cryptocurrencies) when SPP increases.

    Lastly, error correction terms (ECT) in all models appear statistically significant at 1% level with negative sign complying with the ECM theory. Bitcoin model seems to be correcting 23.68% of its previous period disequilibrium in the way converging its long-run level. This adjustment speed in Etherem, Dash, Litcoin, and Monero is 12.76%, 10.20%, 22.91%, and 14.27% respectively.

    The main limitation of the study is latency (novelty, obscurity, and intangibilty) of majority of cryptocurrency related information. This is a brand new market and a brand new topic for academic researches. We also believe that if we could proxy adaptation of cryptocurrency (legalization of any cryptocurrency as a payment tool), we believe it would improve our model further. In addition, few cryptocurrencies comprise speculative bubbles, particularly Bitcoin, thus, future researches should attempt to measure volume of this bubble addressing to question “are we in the peak of Bitcoin bubble?”.


References

Bouoiyour, J., and Selmi, R. (2016). Bitcoin: A beginning of a new phase? Economics Bulletin, 36, 1430–40.

Brill, A., and Keene, L. (2014). Cryptocurrencies: The Next Generation of Terrorist Financing? Defence Against Terrorism Review, 6(1), 7-30.

Chiu, J., and Koeppl, T. (2017). The Economics of Cryptocurrencies - Bitcoin and Beyond. Bank of Canada.

Chu, J. Chan, S., Nadarajah, S., and Osterrieder, J. (2017). GARCH Modelling of Cryptocurrencies. Journal of Risk and Financial Management, 10 (17), 1-15. doi:10.3390/jrfm10040017

Dyhrberg, A. (2016). Hedging capabilities of Bitcoin. Is it the virtual gold? Finance Research Letters, 16, 139–44.

El Bahrawy, A., and Alessandretti, L. (2017). Evolutionary dynamics of the cryptocurrency market. Royal Society Open Science, 4(170623), http://dx.doi.org/10.1098/rsos.170623

Katsiampa, P. (2017). Volatility estimation for Bitcoin: A comparison of GARCH models. Economics Letters, 158, 3–6.

Letra, I.J.S. (2016). What drives cryptocurrency value? A volatility and predictability analysis. Available online: https://www.repository.utl.pt/handle/10400.5/12556 (accessed on 15 December 2017).

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.

Newey, W., and West, K. (1987). A Simple, Positive Semi-definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix. Econometrica, 55(3), 703-08.

Pesaran, M.H., Shin, Y., and Smith, R.J. (2001). Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics, 16(3), 289- 326.

Polasik, M., Piotrowska, A., Wisniewski, T.P., Kotkowski, R., and Lightfoot, G. (2015). Price Fluctuations and the Use of Bitcoin: An Empirical Inquiry. International Journal of Electronic Commerce, 20(1), 9-49

Poyser, O. (2017). Exploring the determinants of Bitcoin’s price: an application of Bayesian Structural Time Series. Dissertation.

Schwarz, G.E. (1978), Estimating the dimension of a model. Annals of Statistics, 6(2), 461–464, doi:10.1214/aos/1176344136.

Sovbetov, Y., and Saka, H. (2018). Does it take two to tango: Interaction between Credit Default Swaps and National Stock Indices. Journal of Economics and Financial Analysis, 2(1), pp. 129-149.

Yermack, D. (2013). Is Bitcoin a real currency? An economic appraisal. NBER Working Paper Series, No: 19747. doi:10.3386/w19747


APPENDIX


Table 1A. Bitcoin Legality by Country and Classification


Name

Bitcoin Legality

Classification


Name

Bitcoin Legality

Classification

1

Afghanistan

Illegal

Currency

59

Lebanon

Legal

No Information

2

Aland Islands

Legal

Currency

60

Liberland

Legal

Currency

3

Algeria

Illegal

Currency

61

Libyan Arab Jamahiriya

Legal

Money

4

American Samoa

Restricted

Commodity

62

Liechtenstein

Legal

Currency

5

Andorra

Neutral / Alegal

No Information

63

Lithuania

Legal

Currency

6

Argentina

Neutral / Alegal

Property

64

Luxembourg

Legal

Currency

7

Australia

Legal

Currency

65

Malaysia

Neutral / Alegal

No Classification

8

Austria

Legal

Currency

66

Maldives

Neutral / Alegal

No Information

9

Azerbaijan

Legal

Currency

67

Malta

Legal

Currency

10

Bangladesh

Illegal

No Information

68

Mauritius

Neutral / Alegal

No Classification

11

Barbados

Neutral / Alegal

No Information

69

Mexico

Restricted

Currency

12

Belarus

Legal

No Information

70

Monaco

Legal

Currency

13

Belgium

Legal

Currency

71

Mongolia

Legal

No Information

14

Bolivia

Illegal

No Information

72

Morocco

Illegal

No Information

15

Brazil

Legal

Commodity

73

Nepal

Restricted

No Classification

16

Brunei Darussalam

Legal

Currency

74

Netherlands

Legal

Commodity

17

Bulgaria

Legal

Currency

75

New Zealand

Legal

Commodity

18

Canada

Legal

Barter Good

76

Nicaragua

Legal

No Information

19

Chile

Legal

No Information

77

Nigeria

Neutral / Alegal

Currency

20

China

Restricted

Commodity

78

Northern Mariana Islands

Legal

Commodity

21

Colombia

Neutral / Alegal

No Classification

79

Norway

Legal

Commodity

22

Congo

Legal

No Information

80

Pakistan

Neutral / Alegal

No Classification

23

Costa Rica

Legal

Currency

81

Paraguay

Neutral / Alegal

No Classification

24

Croatia

Legal

Currency

82

Peru

Neutral / Alegal

No Classification

25

Cuba

Legal

Currency

83

Philippines

Legal

Barter Good

26

Cyprus

Legal

Currency

84

Poland

Legal

Property

27

Czech Republic

Legal

Currency

85

Portugal

Legal

No Classification

28

Denmark

Legal

Currency

86

Republic of Macedonia

Illegal

No Information

29

Ecuador

Illegal

No Information

87

Reunion

Legal

Commodity

30

Egypt

Restricted

Commodity

88

Romania

Legal

Currency

31

Estonia

Legal

Currency

89

Russian Federation

Illegal

Currency

32

Finland

Legal

Currency

90

San Marino

Legal

Currency

33

France

Legal

Commodity

91

Saudi Arabia

Restricted

No Information

34

Gabon

Neutral / Alegal

No Information

92

Serbia

Legal

No Information

35

Georgia

Legal

No Classification

93

Singapore

Legal

Currency

36

Germany

Legal

Barter Good

94

Slovakia

Legal

Currency

37

Greece

Legal

Currency

95

Slovenia

Legal

Currency

38

Hong Kong

Legal

Commodity

96

South Africa

Legal

Currency

39

Hungary

Legal

Currency

97

South Korea

Legal

No Classification

40

Iceland

Legal

Currency

98

Spain

Legal

Currency

41

India

Neutral / Alegal

Commodity

99

Svalbard and Jan Mayen

Legal

Commodity

42

Indonesia

Neutral / Alegal

Commodity

100

Sweden

Legal

Commodity

43

Iran

Legal

No Classification

101

Switzerland

Legal

Currency

44

Iraq

Legal

No Information

102

Taiwan

Legal

No Information

45

Ireland

Legal

Currency

103

Thailand

Legal

Commodity

46

Isle of Man

Legal

No Information

104

Tunisia

Neutral / Alegal

No Classification

47

Israel

Legal

Commodity

105

Turkey

Legal

Commodity

48

Italy

Legal

Currency

106

Ukraine

Legal

Currency

49

Japan

Legal

Currency

107

United Arab Emirates

Legal

Currency

50

Jersey

Legal

Currency

108

United Kingdom

Legal

Currency

51

Jordan

Neutral / Alegal

No Classification

109

United States of America

Legal

Property

52

Kazakhstan

Neutral / Alegal

Currency

110

Uruguay

Neutral / Alegal

Property

53

Kenya

Neutral / Alegal

No Classification

111

Uzbekistan

Legal

Currency

54

Kosovo

Neutral / Alegal

No Information

112

Venezuela

Neutral / Alegal

Commodity

55

Kuwait

Legal

No Information

113

Viet Nam

Neutral / Alegal

Property

56

Kyrgyzstan

Neutral / Alegal

Currency

114

Zambia

Restricted

No Information

57

Latvia

Legal

Currency

115

Zimbabwe

Legal

Commodity



Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.