Volume 119 No. 17 2018, 497-508 ISSN: 1314-3395 (on-line version) url: http://www.acadpubl.eu/hub/ http://www.acadpubl.eu/hub/ HOW GOOD IS THE BITCOIN AS AN ALTERNATIVE ASSET FOR HEDGING? By 1 Dr. HariharaSudhan R., 2 Dr. Sowmya S., 1 Assistant Professor, School of Management, Hindustan Institute of Technology and Science, Chennai 603 103.Tamilnadu, India. 2 Assistant Professor, Indian Institute of Management, Lucknow 1 rhsudan@hindustanuniv.ac.in, kbsudhan@gmail.com 1.Introduction In recent years,the cryptocurrencies have received broader attention and adoption from the investing public. The price appreciation in value of digital currencies has increased to many folds that captivated from individual investors to wall street banks and high frequency traders.an important feature of bitcoin is that they are not backed by any central bank unlike the traditional currencies. The supply of these currencies are limited to a design of protocol (Bouri et al., 2017). The Bitcoin is one of the most popular cryptocurrencies and accounts about 41% of the cryptocurrencies capitalization (Katsiampa,2017).Bitcoin was invented in 2009 by a group of people in the name of SathoshiNakamoto (Briere et al., 2015).Since its introduction, bitcoin has grown steeper and sweeping the investors worldwide. The value of Bitcoin grew from US $ 6 billion (Bouri et al.,2017) in 2015 to US $ 167 billion in November 2017(Source: CoinMarketCap) indicating the tremendous growth. The growing literature in the bitcoin led to the interesting arguments whether the bitcoin is to be treated as commodity or a currency. Bitcoins are treated as a digital gold as they have many similarities. Both these assets are not controlled by any authorities, their supply is limited and they exhibit high price volatility (Dyhrberg, 2016). Glaser et al., (2014) in their study found that majority of users treat bitcoin as a speculative asset rather than the currency. This paper considers the bitcoin as an alternative asset and aims to identify the hedging possibilities of bitcoin against gold, currency and stock market. The study analyses the hedging capabilities of Bitcoinagainst gold, USD- JPY, USD- Euro returns and S&P index The daily closing prices of the assetsfor the period July 2010 to 497
October 2017are considered. The return of the assets is modeled using VAR DCC M GARCH model to capture the time varying conditional correlation between the assets. The results of the study found that long term volatility persistence is less in Bitcoin compared to other assets. Thevolatility of Bitcoin is immune to the rise and fall in the stock market s volatility. The result clearly indicate that bitcoins can be used to hedge against the USD- JPY returns and the stock market. The contribution of the study is in two folds. It is one of the few studies which captures the volatility transmission between the Bitcoin and other assets. The results of study would be of great help to hedge fund managers and portfolio managers. Secondly the period captured in the study includes the period of high and low prices of bitcoins. The rest of the paper is organized as follows. Section 2 presents the methodology. Section 3 describes the data and presents the results. Section 4 concludes. 2. Methodology The Vector Auto Regressive Dynamic Conditional Correlation Multivariate GARCH model is used to estimate the return and volatility spillover between the bitcoin and the other assets considered in the study. The conditional mean equation is modelled using the VAR approach. Let R i,t = (R B t, R S, G, t R t R E t, R J B t ) be the (5x1) vector containing the returns of assets at the time period t. R t represents the returns on Bitcoinindex,R S G t represents the returns on S&P 500 Index, R t represents the returns from gold, R E J t represents the returns of USD- Euro rate and R t represents the returns of USD- JPY rate. The conditional mean equation is specified as follows: Eqn (1) Where R i,t is the return of the asset i,c is the vectorof (5 x1) containing the constant terms of the return equation,ω R ij,t-1 is the VAR coefficientsof lagged own and cross asset returns andε t is a vector of error terms (ε 1, ε 2, ε 3,ε 4,ε 5 ). The conditional meanequationis the function of its own past returns and cross asset past returns. The error term of the mean equation is used to model the conditional volatility. In order to model the conditional volatility and capture the volatility spillover DCC- M GARCH model 498
is used. For all these models VAR GARCH (1,1) is considered as suggested by Ling and McAleer(2003). Eqn (2) where H t is 5 X 5 conditional variance and covariance matrix which includes the time varying variance and conditional covariance. The diagonal elements of H t captures the time varying volatilities of the assets. The non-diagonal elements of H t captures the time varying conditional covariance of the assets. Further the conditional variance and covariance matrix H t is decomposed as H t = D t P t D t Eqn(3) Where D t is the diagonal conditional volatility and P t is timevarying conditional correlation matrix. D t = diag,t Eqn(4) The time varying conditional variances is modeled as Eqn(5) Where is the 5X1 vector of constant terms, A and B are the ARCH and GARCH coefficients. The ARCH coefficients measure the short term persistence and GARCH coefficients measures the long term persistence and volatility clustering. The captures the volatility spillovers between the assets. measures the own conditional ARCH effects. measures the spillover effect of the j on the conditional volatility of i. measures the past volatility effect of the i on the conditional volatility of i. The sum of and should be less than 1 to ensure the long term persistence in the conditional volatility. The Dynamic conditional correlation takes the time varying correlation matrix from P t.the conditional correlation matrix Pt takes the following form Eqn (6) 499
Where, Q t is a 5x 5 positive definite matrix and = ( and given as Eqn (7) Where Q t is the unconditional correlation matrix of standardizedresiduals., are nonnegative scalars and DCC parameters.the DCC parametersare estimated using Quasimaximum likelihood algorithm. The sum of the two DCC parameters should be less than one. The time varying correlation are obtained using Eqn (8) 3.Data The daily price of Bitcoin index, S&P 500 index, Gold, USD- Euro and USD- JPY are considered for the period July 2010 to the October 2017. The period is chosen based on the availability of Bitcoin Index data. The data of Bitcoin Index is obtained from the Coindesk (www.coindesk.com).the prices of S&P 500, USD- Euro and USD- JPY are obtained from Bloomberg. The returns are estimated using the closing daily prices of indices. The returns are calculated as R t = ln(p t /P t-1 ). Table 1: Descriptive statistics ofthe returns USD- JPY Bitcoinreturn S&P return USD-Euro return Gold return return Mean 0.0042 0.0003-3.89E-05 2.39E-05 0.000102 Median 0.0013 0.0000 0.0000 0.0000 0.000000 Maximum 0.4245 0.0463 0.0306 0.04580 0.034854 Minimum -0.4915-0.0690-0.0267-0.095121-0.037820 Std. Dev. 0.0586 0.0075 0.0048 0.008542 0.005050 Skewness -0.3765-0.5217-0.0358-0.775804 0.021242 Kurtosis 15.4835 11.7442 6.6667 13.25487 9.913867 Jarque-Bera 17348.11*** 8601.72*** 1491.95*** 11931.29*** 5302.189*** ADF test -50.3840*** -54.79*** -51.70*** -52.78*** -51.07*** Q(20) 74.4*** 60.6*** 12.42 15.97 15.99 Observations 2662 2662 2662 2662 2662 Notes: T his table represents the descriptive statistics of various assets. ***, ** and * shows the significance at 1%, 5% and 10% respectively.adf test is the Augmented Dickey fuller test for testing stationarity of the series. Q(20) is the Ljung-box Q statistic for the return series to test significant serial correlation. 500
Table 1 presents the descriptive statistics of returns of Bitcoin, S&P 500 Index, USD- Euro rate, Gold and USD- JPY rate. Bitcoin offers the highest average daily return followed by the investment in equity while USD- Euro rates exhibit negative return. The Bitcoin is the most volatile asset with the standard deviation of 5.86%. This is evident from the fact that high returns leads to high risk. The asset returns appear to be non-normal and exhibit negative skewness except the USD- JPY return. The stationarity of the return series is found using the Augmented Dickey Fuller test and the results indicate that the returns are stationary at the level with the significance of 1%. There is no significant auto correlation in the return series of USD- Euro return, Gold return and USD- JPY return. Fig 1 : Price and Return Plot of Bitcoin 7000 6000 5000 4000 3000 2000 1000 Price and Return Plot - Bitcoin 0.6 0.4 0.2 0-0.2-0.4 0-0.6 7/18/2010 7/18/2011 7/18/2012 7/18/2013 7/18/2014 7/18/2015 7/18/2016 7/18/2017 Bitcoin Index Bitcoin Return Fig 1 graphs the price and the return of the bitcoin index. The figure clearly indicates that there is a steady increase in the prices of bitcoin from 2013. However, from 2015 the increase in the prices of bitcoin is very steep resulting in considering the price increase as a speculative bubble. Table 2: Unconditional correlation between the returns Bitcoin Return Gold return S&P Return Bitcoin Return 1 Gold return 0.0098 1 S&p 0.0116-0.0067 1 USD-Euro return USD- JPY Return 501
Return Usd-euro return 0.0134 0.2712 0.1682 1 USD- JPY Return -0.0115-0.3349 0.2563-0.2433 1 Notes: This table represents the unconditional correlation between the asset returns. Table 2 presents the unconditional correlation between the asset returns. The results indicate that there existsa negative correlation between the Bitcoin and USD- JPY return. The Bitcoin exhibits positive correlation with Gold, S&P index and USD- Euro rate return for the sample period. The Gold return exhibit negative correlation between S&P return and USD- JPY return indicating the hedging opportunities. The similar views are found in the study of Capieet al., (2005). Their study found that Gold can be used as hedge against dollar currencies as they are not controlled by same institutions. Similar relationship is found between the Bitcoin and USD- JPY return. Table 3: Estimation Results of DCC M GARCH Method Panel A : Mean Equation Constant 0.00239*** Bitcoin Gold S&P (0.000) Bitcoin t-1 0.057** (0.022) Gold t-1 0.2248*** (0.069) S&P 0.2868 Return t-1 (0.193) USD-Euro 0.0143 return t-1 (0.111) USD- JPY 0.1908 Return t-1 (0.1600) -0.0000 (0.0001) 0.0022 (0.0017) -0.0241 (0.02) 0.0070 (0.018) -0.0932*** (0.0270) -0.0901*** (0.0291) Return 0.0005*** (0.0001) 0.0037** (0.0018) -0.0291** (0.0131) -0.0265 (0.0204) -0.0435* (0.0232) -0.0459** (0.0228) USD-Euro return -0.00004 (0.00008) 0.0035*** (0.0013) 0.0338*** (0.0088) 0.0338*** (0.0117) -0.0616*** (0.0193) -0.0676*** (0.0163) USD- JPY Return 0.0001* (0.0000) -0.0007 (0.0011) -0.0137 (0.0105) 0.0251** (0.0111) 0.0193 (0.0174) -0.0236 (0.0200) Panel B : Variance Equation Bitcoin(i) Gold(i) S&P Return(i) Constant 0.000 0.000 0.000 (0.000) (0.000) (0.000) ARCH Coefficients A i,j USD-Euro return(i) 0.000 (0.000) USD- JPY Return(i) 0.000 (0.000) 502
Bitcoin(j) 0.230*** 0.0028*** -0.00183 0.000998** 0.0000 (0.02) (0.0007) (0.0014) (0.0004) (0.0008) Gold(j) -0.4965*** 0.0255*** -0.0115** -0.0015 0.0061* {0.1008) (0.0041) (0.0054) (0.0018) (0.0037) S&P 0.1678-0.262*** 0.097*** -0.000645 0.0143*** Return(j) {0.1638} (0.0040) (0.011) {0.001282} (0.0033) USD-Euro -0.1709 0.0193*** 0.0123 0.0148*** -0.0031 return(j) {0.1791} (0.0073) (0.0125) (0.0024) (0.0041) USD- JPY -0.3646** 0.0356*** -0.0362*** -0.0014 0.0281*** Return(j) {0.1762} (0.0060) (0.0120) (0.0025) (0.0039) GARCH 0.773*** 0.9693*** 0.8806*** 0.9829*** 0.9589*** Coefficients (0.0145) (0.0047) (0.012) (0.0026) (0.0045) B i,i Persistence 1 0.9948 0.9776 0.998 0.987 Panel C : Multivariate DCC Equation DCC(1) 0.016 *** (0.00170 DCC(2) 0.977*** (0.003) Panel D : Multivariate Ljung-Box test Q(20) 991.74 Q(40) 508.50 Notes: T his table represents the estimation results of VAR- DCC MGARCH model.***, ** and * shows the significance at 1%, 5% and 10% respectively. Panel A represents the mean equation of all the five assets considered in the study. Panel B represents the variance equation. ARCH coefficients captures the spillover effects between the assets. T he j the term representsspillover effect of the asset (j) on the conditional volatility of asset (i). Persistence term is the sum of ARCH and GARCH coefficients of the assets. Persistence of Bitcoin is (0.230+0.773). Q (20) represents the Multivariate Ljung Box test statistic with lags 20 and 40. Table 3 reports the estimation result of VAR-DCC MGARCH Model. The Panel A exhibits the meanequation obtainedusing VAR.The auto regressive term of Bitcoin return is found to be positive and significant. Thisresult clearly emphasis that the past returns of Bitcoin can be used to predict the future returns. However, the autoregressive term of Gold, S&P index and USD- JPY rate are found to be negative and insignificant. Thus the past returns of these assets does not predictthe future returns. Maghyerehet al., (2017) alsofound the similar finding that the past return of the gold is negative and insignificant with the current returns. The autoregressive term of USD- Euro returnis found to be negative and significant. 503
The increase in the gold return increases bitcoin return.this indicates that the gold return influences the bitcoin return. However, the past Bitcoin return has a significant and positive impact on the S&P return and the USD- Euro Return. The increase in the USD- JPY and USD Euro rate return reduces thegold returnsignificantly. Thus gold has the hedging capabilities for the forex returns. The increase in the gold and currency return decreases the S&P return. This is because the investing community treats gold and currency returns as alternative for the S&P returns. The panel B in table 3 exhibits the variance equation. The ARCH coefficients represents the short term dependence and GARCH coefficient represents the volatility persistence. The results exhibit highlysignificant long term persistence in the volatility across the asset classes. The long term persistence is less in bitcoin compared to the other assets. This result indicates that the shocks has lesser long term effect on the volatility of bitcoin compared to other assets classes. It is worth to note that the volatility of Bitcoin is not influenced or impacted by the rise or fall in the volatility of stock returns. It s the Gold and Forex rate which affects the volatility of the Bitcoin. The gold and USD- JPY volatility has negative effect on the bitcoin. The panel Cof table 3 indicates the DCC parameters (θ 1, θ 2 ). The two DCC parameters are found be positive and significant. The positive and significant DCC parameters indicate the substantial time varying co-movements between the asset class. The sum of the DCC parameters is less than 1 exhibits the stability of the DCC M GARCH model. Fig 2: Time varying conditional correlation between Bitcoin and Gold 0.25 Conditional Correlation of Bitcoin with Gold 0.20 0.15 0.10 0.05-0.00-0.05-0.10-0.15-0.20 2010 2011 2012 2013 2014 2015 2016 2017 504
Fig 3: Time varying conditional correlation between Bitcoin and S&P Index 0.3 Conditional Correlation of Bitcoin with S&P Index 0.2 0.1 0.0-0.1-0.2-0.3 2010 2011 2012 2013 2014 2015 2016 2017 Fig 4: Time varying conditional correlation between Bitcoin and USD- Euro Exchange Rate 0.3 Conditional Correlation of Bitcoin with USDEuro Exchange rate 0.2 0.1 0.0-0.1-0.2-0.3 2010 2011 2012 2013 2014 2015 2016 2017 Fig 5: Time varying conditional correlation between Bitcoin and USD- JPY Exchange Rate 0.20 Conditional Correlation of Bitcoin with USDJPY exchange rate 0.15 0.10 0.05-0.00-0.05-0.10-0.15-0.20 2010 2011 2012 2013 2014 2015 2016 2017 505
Table 4: Average Dynamic Conditional Correlation Mean t-statistic Bitcoin and Gold 0.0088 7.28*** Bitcoin and S&P Index 0.0050 3.79*** Bitcoin and USD Euro 0.0068 5.36*** Bitcoin and USD JPY -0.0109-10.31*** Notes: T his table represents the average conditional correlation estimates. ***, ** and * shows the significance at 1%, 5% and 10% respectively. Table 4 reports the average conditional correlation and the t statistics for the mean test. The time varying conditional correlation of the DCC MGARCH model is extracted and the average pairwise conditional correlation is calculated for the Bitcoin and other asset class. The mean test is estimated to find whether the average conditional correlation is different from zero or not. The results indicate Bitcoin exhibits positive correlation between Gold, S&P index, USD- Euro returns. However, the correlation with the stock market is very less indicating the chances of hedging opportunities. The Fig 4 exhibits that after 2013, the time varying correlation between the stock market and bitcoin are found to be negative and sometimes with very less positive correlation. The finding is in line with the study of Dyhrberg (2016), who also found that bitcoins has hedging capabilities against the FTSE stock index. The time varying correlation between the Bitcoin and USD- JPY is found to be negative. This is because among the G7 countries, Japan is the only country that accepted the bitcoin as a legitimate currency. The bitcoin trade in Japan accounts around 50% of the global trade volume. The result clearly indicate that Bitcoin can be used for hedging against USD- JPY returns and stock market. Conclusion The study examined hedging possibilities of bitcoin against Gold, S&P index, USD- Euro and USD- JPY rates. The result of the study found that the Bitcoin exhibit lesser long term volatility persistence compared to other assets. The Bitcoin is immune to the stock market volatility. It is also found that there existsa negative conditional correlation between the bitcoin and USD-JPY rates indicating the capabilities of hedging opportunities. Thus bitcoin has a clear place for hedging opportunities against the stock market and JPY returns. 506
References 1. E. Bouri et al., Does Bitcoin hedge global uncertainty? Evidence from wavelet-based quantile - quantile regressions, Finance Research Letters (2017) 2. E. Bouri et al - 3. Anne HauboDyhrberg, Bitcoin, Gold and the Dollar A GARCH Volatility analysis, Finance Research Letters (2016), 16, 85 92. 4. Anne HauboDyhrberg, Hedging capabilities of Bitcoin. Is it a virtual gold?, Finance Research Letters (2016), 16, 139-144 5. Aurelio F. Bariviera, The inefficiency of Bitcoin revisited: A dynamic approach, Economic Letters (2017), 161, 1-4 6. Paraskevikatsiampa, Volatility estimation of Bitcoin : A - 7. Comment [SS1]: Full author name. no et al should be used in references 8. Capie et al.,(2005 Gold as the hedge against the dollars 9. K. Jayalakshmi, K. Hari Babu (2017), Alternator Ideals In Vinberg (-1, 1) Rings, Inter National Journal Of Pure And Applied Mathematics, 113 (6) : 38 46 10. Dirk G. Baur, Thomas K.J. McDermott, Why is gold a safe haven(2016) Journal of behavioural and experimental Finance. 507
508