See discussions, stats, and author profiles for this publication at: [Link]

net/publication/330545680

DYNAMIC LINKAGES AMONG BITCOIN, GOLD PRICES AND EXCHANGE RATES

OF US DOLLAR IN JPY, GBP AND CNY: DCC EGARCH APPROACH

Article · January 2019

CITATIONS READS

3 215

2 authors, including:

Wajdi Moussa
Institut Supérieur de Gestion de Tunis
13 PUBLICATIONS 5 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Can we reckon Bitcoin as a hedge for American Dollars? View project

child developement View project

All content following this page was uploaded by Wajdi Moussa on 22 January 2019.

The user has requested enhancement of the downloaded file.

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS

DYNAMIC LINKAGES AMONG BITCOIN, GOLD

PRICES AND EXCHANGE RATES OF US DOLLAR IN
JPY, GBP AND CNY: DCC EGARCH APPROACH

HAZGUI SAMAH1
University of Tunis
hazguisamah066219@[Link]
MOUSSA WAJDI2
University of Tunis
[Link]@[Link]
RYM REGAÏEG3
University of Tunis
[Link]@[Link]

Abstract

The aim of this paper is to examine the relationship between bitcoin, exchange rates of US
Dollar in JPY, GBP and CNY and gold prices. The daily data for the period of 19/07/2010
to 02/05/2018 are used to model volatilities between our variables with univariate and
multivariate GARCH models (EGARCH, DCC-EGARCH). The results show that bitcoin
has the benefits of both commodities and currencies in the financial markets and it is useful
for portfolio and risk management.

Keywords: Bitcoin market, Gold price, Exchange rate EGARCH, DCC-EGARCH

JEL classification: D74, G14, G1, D24

1. INTRODUCTION

Since its creation in 2009, there is a huge interest in research addressing the
economic-financial aspects of Bitcoin which evoked a debate on whether it is a
currency or a commodity. As we know, Bitcoin is characterized by high levels of
return and volatility which could affect other assets (European Central Bank,

1
University of Tunis, École Supérieure de Commerce de Tunis
2
University of Tunis, Institut Supérieur de Gestion de Tunis, Laboratoire : Gouvernance
d'Entreprise, Finance Appliquée et Audit (GEF2A)
3
University of Tunis, Institut Supérieur de Gestion de Tunis, Laboratoire : Gouvernance
d'Entreprise, Finance Appliquée et Audit (GEF2A)
VOLUME 10 NUMBER 2 JULY 2018 239
 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
2012). Furthermore, a very little research has been written about Bitcoin and
commodity, in general, and Bitcoin and gold, in particular. For example, Yermarck
(2015) shows that Bitcoin is more like a speculative investment than a currency;
whereas Polasik et al. (2015) show that Bitcoin is a medium of exchange.
Monaghan (2018) consider Bitcoin as a bubble waiting to burst. Luther and Salter
2017 indicate that this cryptocurrency would be seen as an alternative to traditional
stores of values, such as gold, and will be considered as a digital gold (Popper
2015). Baek and Elbeck (2015) found that Bitcoin returns are not influenced by
fundamental economic factors, but by their investors. Brière et al. (2015) show the
weak correlation of Bitcoin with alternative investments (commodities and hedge
funds) and traditional assets (stocks, bonds, currencies) and show the
diversification capabilities of Bitcoin despite its high volatility. Ji et al. (2017)
show a low correlation between Bitcoin and gold market. Bouri et al. (2017) show
that Bitcoin is a diversification for major world index, bonds, oil, gold and for
American dollar. Li and Wang (2017) found a significant relationship in the short
and long terms between the Bitcoin price and changes in economic fundamentals.
Dyhrberg 2016 proves that Bitcoin has comparable hedging capabilities and safe
havens like gold, and it would be categorized between gold and American dollar.
Baur et al. (2018) criticize the paper of Dyhrberg (2016) and found that Bitcoin has
different characteristics to gold and American dollars.
This review is structured as follows: Section 2 describes the data and the
methodology and section 3 discusses the results of the estimated models which are
followed by concluding remarks in section 4.

2. DATA AND METHODOLOGY FRAMEWORK

The daily data series used for this paper include 2844 observations from
19/7/2010 to 02/05/2018. The starting date is depicted by the accessibility of
Bitcoin prices. The closing prices for the bitcoin index are sourced from
[Link]. USD/JPY, USD/GBP and USD/CNY exchange rates are from
Bloomberg data, gold prices are collected from the World Gold Council. For each
data series, the daily returns are calculated as log (𝑃𝑡 )- log(𝑃𝑡−1) where 𝑃𝑡 is the
daily closing price. Like Dyherberg (2016) study, we assume zero returns for
Saturdays and Sundays to align all series with the Bitcoin data. In this paper, we
use the EGARCH model of Nelson (1991) combined with DCC model to
investigate the significance of asymmetry between the BTC, gold and American
dollar. As known, one of the advantages of this model is that it allows engaging the
dynamics of volatility, the volatility spill overs, the potential asymmetric effect of
shock transmissions and the conditional correlations between series.
We apply in the first step the E-GARCH model to account for volatility
feedback and leverage effects. According to Cappiello et al. (2006) the choice of
univariate model will not affect the sign of the standardized residual and the
correlations would be relatively insensitive to the GARCH model specification.
The mean and variance equations take the following form:

240 VOLUME 10 NUMBER 2 JULY 2018

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
Ret-BTCt = 𝛽0 +𝛽1 Ret-BTCt-1+𝛽2 GOLDt-1+𝛽3 USDEGBPt-1+𝛽4 USDJPYt-
1+𝛽5 USDCNYt-1 +𝜀𝑡 (1)
ln (𝜎𝑡2 )=𝜆0 + 𝜆1 GOLDt-1+𝜆2 USDGBPt-1+𝜆3 USDJPYt-1+𝜆4 USDCNYt-1+α

(𝜀𝑡−1 /𝜎𝑡−1)+ γ ( |𝜀𝑡−1/𝜎𝑡−1| -√2⁄𝜋 )+ δ ln𝜎𝑡−1

2
(2)

Where 𝜀𝑡−1 denotes the previous period’s squared residual series, 𝛼

represents a magnitude effect or the symmetric effect of the model, δ measures the
persistence in conditional volatility irrespective of anything happening in the
market. The volatility takes a long time to die out following a crisis in the market
when δ is large. The ꙋ coefficient measures the leverage effect. If ꙋ<0 the good
news generate less volatility than bad news and if ꙋ≠0, the impact is asymmetric.
Then, in a second step the standardized residuals are used to estimate the
time varying correlations and the DCC model as formulated by Engle (2002) is
estimated as :
𝐻𝑡 = 𝐷𝑡 𝑅𝑡 𝐷𝑡 (3)
1\2 1\2 1\2 1\2
Where 𝐷𝑡 =diag (ℎ1.𝑡 , … , ℎ𝑛.𝑡 ), 𝑅𝑡 =diag( 𝑞1.𝑡 , … , 𝑞𝑛.𝑡 )𝑄𝑡 (
−1\2 −1\2
𝑞1.𝑡 , … , 𝑞𝑛.𝑡 ), 𝐻𝑡 is an n×n conditional covariance matrix, 𝑅𝑡 is the
conditional correlation matrix and 𝐷𝑡 is the diagonal matrix with time varying
standard deviations on the diagonal.
The dynamic conditional correlation structure is modelled as follow:
𝑄𝑡 = (1 − 𝜃1 − 𝜃2 )𝑄̅ + 𝜃1 𝑍𝑡 𝑍𝑡′ +𝜃2 𝑄𝑡−1 (4)
Where, 𝑄𝑡 is the unconditional variance between series, 𝑄̅ is the
unconditional covariance between the univariate series estimated in first step,
𝜃1 , 𝜃2 are non-negative scalar parameters satisfying 𝜃1 + 𝜃2 < 1.
To determine the dynamic correlations between our variables, EGARCH
and DCCEGARCH models are estimated with maximum likelihood estimation
procedures. The likelihood function is maximized by Broyden, Fletcher, Goldfarb,
Shanno (BFGS) numerical algorithms to estimate the parameters.

3. RESULTS AND DISCUSSION

This section presents all of the statistical procedures for the estimates of
volatility measures according to the models presented in section 2. The results
reported in Table 1 shows that both mean and volatility of the Bitcoin is greater
than the other series suggesting that its return exhibit high volatility clustering with
a standard deviation of 0.05867 following with gold by 0.008309, while CNY
exhibit the very low volatility clustering. All return series are skewed negatively,
except for the USD/GBP and USD/CNY. All series have kurtosis value in excess
of that in a normal distribution. Table 1 also shows the results of unit root tests

VOLUME 10 NUMBER 2 JULY 2018 241

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
over the return series. We perform the Augmented Dickey fuller (ADF) and
Phillips Perron (PP) tests to check the property of the data series and in all cases we
reject the null hypothesis of unit roots. Thus, all variables are stationary. This is
also evident from the time series plot of return series which is presented in Fig 1.
Table 2 reports the correlation coefficients of stock returns for period from July 19,
2010 to May 2, 2018. Across all of the asset returns, BTC correlates positively with
other series except the gold. Note that the low or negative correlations between
assets are desirable from a risk management perspective. The result from
EGARCH model has reported in Table 3. From the mean equation it is confirmed
that GOLD has positive impact on BTC at 1% level of significance suggesting that
10% increase in the GOLD leads to 21,49 % increase of BTC price. The coefficient
on the exchange rates suggests that bitcoin returns are more sensitive to the value
of the yen. Therefore, regional or country specific effects are presented. Moreover,
this finding is similar to previous gold results (see Tully and Lucey 2007) and
indicates that BTC may also be useful in hedging against the dollar like gold.
Coming to the variance equations which indicate that positive volatility shock to
the sterling and Yen exchange rates decreases the variance of the BTC returns
which indicate that this cryptocurrency is a relatively safe asset in such situations.
Thus, as already noted by Dyhrberg (2016), BTC may have some risk management
capabilities like gold against the dollar (see Capie et al. 2005 results). Additionally,
the estimated asymmetric term (γ) is positive and statistically significant
confirming that good and bad news have not an asymmetric impact on the BTC
volatility like gold (see Hammoudeh and Yuan 2008 results). Besides δ is positive
and relatively large above 0.9, hence volatility takes a long time to die out. Added
to this, table 3 displays weak levels of persistence (δ + α < 1) implying that the
daily return series are stationary and absent of volatility clustering or market
momentum. Lastly, the statistical significance of all the coefficients indicates the
presence of conditional heteroskedasticity in the daily return series. As positive or
negative shocks do not affect BTC and gold returns, it would be able to be used as
a hedge market risks which affect other assets asymmetrically.
The second stage of the estimation uses the standardized residuals obtained
from the E-GARCH univariate model to estimate the time-varying DCC
correlations. Table 4 indicates that the time-varying correlations are mean reverting
since 𝜃1 +𝜃2 <1. The coefficient 𝜃1 measures the effect of past standardized
innovations on dynamic conditional correlations, while 𝜃2 reports the impact of
lagged dynamic conditional correlations on the current dynamic conditional
correlations. In addition, these parameters are significant, indicating significant
variation over the specified period. More specifically, the statistical significance of
𝜃1 and 𝜃2 indicates that a DCC model is suitable to be used. The estimated
conditional correlations between variables shown in table 4 are very similar to the
value recorded in table 3. The low values of 𝜃1 and the high values of 𝜃2 indicate
that the correlation process is resistant to shocks and reverts to the mean quickly.
This indicates that the correlations amongst the variables should be stable without
many outliers.

242 VOLUME 10 NUMBER 2 JULY 2018

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
4. CONCLUSION

The objective of this paper was to investigate the dynamic relationship

between BTC, gold and American dollar from July 2010 to May 2018. The linear
models fail to check the dynamics between such series due to serial correlation and
non-normal distribution. Therefore, this paper investigates the time varying
relationship between series using GARCH framework. The empirical results
estimation reveals an interesting finding which has important implications for
investors, portfolio managers and policymakers. Our finding contradicts with
Kristoufek (2015) and Bouoiyour and Selmi (2015) who find insignificant
relationship between Bitcoin and gold prices. However, they are in line with Li and
Wang (2017) who show that Bitcoin is sensitive to macroeconomic indicators. This
review shows that bitcoin is similar to gold and dollar and as bitcoin is
decentralized it will never behave exactly like them on the market. Also, this
cryptocurrency is similar to gold as it reacts to similar variables in the GARCH
models, has similar hedging capabilities and similar react to news. However, the
frequency is higher for bitcoin as its trading is faster and investors’ reactions are
quicker. To conclude, Bitcoin can combine some of the advantages of both in the
financial market. Following Dyhrberg (2016) findings bitcoin would be between
gold and American dollar. For future research we will be focus on the Bitcoin-
stock nexus in China and Japan, which have a large part of Bitcoin trading
activities and the inclusion of other cryptocurrencies will be an interesting addition
to the current study.

REFERENCES

Baur, D.G., Dimpfl, T., Kuck, K. (2018), Bitcoin, gold and the US dollar – A
replication and extension. Finance Research Letters, 25, DOI: 10.1016/
[Link].2017.10.012
Baek. C, Elbeck. M, (2015), Bitcoin as an investment or speculative vehicle? A
first look. Applied Economics Letters, 22, 30-34.
Benjamin M. Blau, (2018), Price dynamics and speculative trading in Bitcoin,
Research in International Business and Finance, 43, 15–21.
Bouri.E, Molnár.P, Azzi.G, Roubaud.D, Hagfors.I, (2017), On the hedge and safe
haven properties of Bitcoin: Is it really more than a diversifier?, Finance
Research Letters, 20, 192–198.
Bouri.E, Jalkh.N, Molnár.P, Roubaud.D, (2017), Bitcoin for energy commodities
before and after the December 2013 crash: diversifier, hedge or safe haven?,
Applied Economics, Forthcoming
Brière.M, Oosterlinck.K, Szafarz.A, (2015), Virtual currency, tangible return:
Portfolio diversification with bitcoin, Journal of Asset Management, 16, 365-
373.

VOLUME 10 NUMBER 2 JULY 2018 243

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
Capie, F, Mills, T.C, Wood, G, (2005). Gold as a Hedge against the dollar. J. Int.
Financ. Mark., Inst. Money, 15, 343–352.
Dyhrberg.A.H, (2016), Bitcoin, gold and the dollar - A GARCH volatility analysis,
Finance Research Letters, 16, 85–92.
Engle.R.F, Ndong.V.K, (1993), Measuring and testing the impact of news on
volatility, Journal of Finance, 48, 1749–1778.
Engle.R, (2002), Dynamic conditional correlation: A simple class of multivariate
generalized autoregressive conditional heteroskedasticity models. Journal of
Business & Economic Statistics, 20, 339-350.
Ji.Q, Bouri.E, Gupta.R, Roubaud.D, (2018). Network Causality Structures among
Bitcoin and other Financial Assets: A Directed Acyclic Graph Approach. The
Quarterly Review of Economics and Finance May 2018,
[Link]
Hammoudeh, S., Yuan, Y, (2008). Metal volatility in presence of oil and interest
rate shocks. Energy Econ. 30, 606–620.
Luther.W. J., Alexander. W. S, (2017). Bitcoin and the Bailout. The Quarterly
Review of Economics and Finance, 66, 50-56
Guesmi.K, Saadi.S, Adid.I, Ftiti.Z, (2018), Portfolio diversification with virtual
currency: evidence from Bitcoin, International Review of Financial Analysis,
[Link]
Katsiampa. P, (2017), Volatility estimation for bitcoin: A comparison of Garch
models. Economics Letters, 158, 3–6.
Koutmos.D, (2018), Bitcoin returns and transactions activity, Economic Letters,
167, 81-85.
Li.X, Wang.C.A, (2017), The Technology and Economic Determinants of
Cryptocurrency Exchange Rates: The Case of Bitcoin. Decision Support
Systems, 95, 49-60
Monaghan. A, (2018), Bitcoin biggest bubble in history, says economist who
predicted 2008 crash. The Guardian (February 02, 2018).
Nelson. D. B, (1991), Conditional heteroskedasticity in asset returns: A new
approach. Econometrica, 59, 347-370
Polasik, M, Piotrowska, A, Wisniewski, T, Kotkowski, R, Lightfoot, G, (2015),
Price fluctuations and the use of Bitcoin: an empirical inquiry, Int. J. Electron.
Commer, 20, 9–49.
Popper. N, (2015), Digital Gold: The Untold Story of Bitcoin, Penguin London.
Tully, E, Lucey, B, (2007). A power GARCH examination of the gold market. Res.
Int. Bus. Financ. 21, 316–325.
Yermack.D, (2015), Is Bitcoin a real currency? An economic appraisal. In D. L. K.
Chuen (Ed.), Handbook of digital currencies (31–44).

244 VOLUME 10 NUMBER 2 JULY 2018

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
APPENDIX
Table 1: Summary Statistics and Stationarity Analysis

Statistics Ret-BTC Ret-GOLD Ret-JPY Ret-GBP

Ret-CNY
MEAN 0.004095 0.0000555 0.0000838 0.0000486 -
0.0000239
S.D 0.05867 0.008309 0.005029 0.004656
0.001329
S.K -0.352273 -0.782166 -0.068235 2.100088
0.569979
K.R 14.65598 14.93389 10.35179 42.83131
25.30690
MIN -0.491528 -0.089128 -0.037722 -0.09962 -
0.011890
MAX 0.424580 0.050705 0.034639 0.084006
0.018382
J.B 16158.47 17166.50 6406.990 190094.7
59119.31
Unit root tests
ADF(none) -8.141139*** -54.09615*** -53.23969*** -12.86731*** -
53.20839***
(0.0000) (0.0000) (0.0000) (0.0000)
(0.0001)
ADF(Constant) -11.92291*** -54.09736*** -53.24888*** -12.88867*** -
53.20427***
(0.0000) (0.0001) (0.0001) (0.0000)
(0.0001)
ADF(Constant +Trend) -11.99625*** -54.09615*** -53.23969 -12.91913*** -
38.60425***
(0.0000) (0.0000) (0.0000) (0.0000)
(0.0000)
PP(none) -52.23889*** -54.10534*** -53.24347***-53.21315 -
49.45542***
(0.0001) (0.0001) (0.0001) (0.0001)
(0.0001)
PP(Constant) -52.20812*** -54.09809*** -53.24982*** -53.20915*** -
49.45471***
(0.0001) (0.0001) (0.0001) (0.0001)
(0.0001)
PP (Constant and Trend) -52.22648*** -54.09716*** -53.24056*** -53.21169*** -
49.45665***
(0.0000) (0.0000) (0.0000) (0.0000)
(0.0000)
Note: S.D: Standard Deviation, S.K: Skewness, K.R: Kurtosis, J.B: Jarque- Berra, in the
parenthesis we report p-values, *** denotes significance at 1% , 5% and 10% levels of
significance.

VOLUME 10 NUMBER 2 JULY 2018 245

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS
Table 2: Correlation Matrix
Ret-BTC Ret-GOLD Ret-GBP Ret-JPY
Ret-CNY
Ret-BTC 1
Ret-GOLD -0.001805 1
Ret-GBP 0.012593 -0.005332 1
Ret-JPY 0.018406 -0.037158 0.137312 1
Ret-CNY 0.029419 -0.016980 0.174006 0.086009 1

246 VOLUME 10 NUMBER 2 JULY 2018

 JOURNAL OF ACADEMIC RESEARCH IN ECONOMICS

Figure 1. BTC, gold and exchange rates returns

Table 3: Exponential Garch(1, 1) dependent variable return on bitcoin
Mean equation Variance equation
GOLDt–1 0.214873*** 2.035436**
(0.046523) (1.116042)
USD-GBP exchange ratet–1 0.159052 -15.02680***
(0.119182) (1.682186)
USD-JPY exchange ratet–1 -0.064496 -14.75246***
(0.095588) (1.885432)
USD-CNY exchange ratet–1 0.135846 29.09949***
(0.432300) (5.540160)
[Link] 0.043841***
(0.021307)
[Link] α 0.018349***
(0.008186)
L.earch_a γ 0.391582***
(0.012524)
[Link] 0.941722***
(0.002930)
Constant 0.004323*** -0.620183***
(0.000403) (0.021203)

Note: *** indicates significance at 1% significance level , T-staistics in parentheses

Table 4: Multivariate Garch Parameter Estimates
Parameters DCC-EGARCH
𝜽𝟏 0.005732*** (0.001394)
𝜽𝟐 0.978194*** (0.008262)
Loglikelihood 52280.02
Akaike info criterion -36.74263
Hannan-Quinn -36.71847
Note: *** indicates significance at 1% significance level , T-staistics in parentheses

VOLUME 10 NUMBER 2 JULY 2018 247

View publication stats