IRDF 2022, Universitas Muhammadiyah Yogyakarta

February, 8th 2022

Is Bitcoin a Better
Portfolio Diversifier than
Gold? A copula and sectoral analysis for China
Prof. Wing Keung Wong
Asia University Taiwan
Outlines
01 Introduction

02 Literature Review

03 Data and Methodology

04 Results and Discussions

05 Three Robustness Test

06 Conclusion
Introduction
Introduction
During the last years, Bitcoin and cryptocurrency have become increasingly
01 important in the financial system, not only as a payment tool but also as a financial
asset (Dyhrberg, 2016).

02 In China, there has been a prohibition from the government to trade

Bitcoin

03 Based on a survey of 100.000 individuals in China, the report

shows that Bitcoin has been considered as one of the ten
favorite investment assets of the middle class in China in 2018
while traders and exchanges try to find ways to bypass the
prohibition by moving the servers out of China
(2018 White Paper on the New Middle Class Survey –
By Prof Xiaobo Wu)
Fig 1. Investment assets by the New Middle Class in China in 2018

10%
Of new middle class in
China has been investing
in cryptocurrencies

In this list, gold also appears as

one of the favorite investments in
China, as part of the precious
metals section. However, gold
and Bitcoin are profoundly
different. Gold is a real and physical
asset and was used as a real
currency during centuries under the
gold standard.
 Bitcoin
On the other hand, Bitcoin is a virtual and “young” asset, created in 2008 by a group of
programmers under the pseudonym Satoshi Nakamoto (Cheah and Fry, 2015).

Bitcoin is only a virtual money derived from mathematical

cryptography and conceived as an alternative to government-backed
currencies. It was originally envisaged that its construction and
digital “mining” processes would mean that Bitcoin prices should
be relatively stable.

Now, it is considered more as a speculative asset than a payment tool

(Horra, 2019). The price volatility and the potential for profit, together
with the prohibition, make investments in cryptocurrencies attractive and
desirable in China.
Study Objectives

the objective of our study is to test whether Bitcoin is really a profitable

investment for Chinese investors, who also invest in stocks and bonds (as
indicated in Figure 1).

Since the part of foreign assets invested by Chinese investors is very small (see
Figure 1), it is reasonably realistic to consider only Chinese stocks and bonds in the
present study.
Research Contribution

In addition to the use of a novel method to estimate the joint

distribution function based on a copula,
our study also contributes to the literature by investigating the
specific case of China in which both Bitcoin and gold repre-
sent specific asset classes

Furthermore, gold has an important cultural aspect in China

because it is used as gift in the new year and wedding seasons
(Cheng, 2014; Hoang et al., 2018b).
Thus, the comparison between these two different alternative
assets in Chinese stock and bond portfolios would provide
important information to Chinese investors and policy makers.
Literature Review
In this literature review, we present a
synthesis of recent studies about
Bitcoin, cryptocurrencies and their
relationship with traditional assets
such as stocks and bonds or with
other alternative assets such as gold.

We notice a huge increase of the number of studies on

Bitcoin and cryptocurrencies since 2015 and mostly from
2017 following the peak of Bitcoin prices on 18
December 2017 at $18,674.
 Academic studies on Bitcoin and cryptocurrencies
can be divided into four different groups

01 The behavior of Bitcoin and cryptocurrencies’ prices

The comparison between Bitcoin/cryptocurrencies and traditional assets

02 or other alternative assets (stocks, bonds, commodities, gold,
currencies, etc.);

03 The hedging ability of Bitcoin and cryptocurrencies in a diversified

portfolio;

The role of Bitcoin and cryptocurrencies in financial innovations,

04 payment systems and computer science
Group 1 behavior of Bitcoin and cryptocurrencies’ prices

Some characteristics can be drawn, such as

speculative bubbles, informational inefficiency,
predictability, day-of-the-week effects,
interdependence of cryptocurrencies, high
volatility, dependence on investors’ sentiment,
dependence on some macroeconomic factors
and on other financial assets, etc.
Group 2 The comparison between Bitcoin/cryptocurrencies and
traditional assets

Table 1 further shows that the volatility of Bitcoin is

higher than that of other assets such as gold, stocks, and
currencies. Moreover, Bitcoin has a low correlation with
traditional asset classes such as stocks, bonds,
commodities, and the USD.

In addition, this relation is asymmetric and nonlinear,

and cryptocurrencies can Granger cause commodity
futures.

Compared to gold, Bitcoin is a less efficient hedge and

safe haven asset for stocks.
 Group 3 The hedging ability of Bitcoin and cryptocurrencies in a
diversified portfolio;

As for the role of Bitcoin in the diversification of portfolios, which is directly related to our study, Group 3 in Table 1
shows that there have been inconclusive results regarding the ability of Bitcoin to be a hedge and a safe haven
asset.

Dyhrberg (2016) found that Bitcoin can be used as a hedge against stocks in the
FTSE index and against the USD in the short run, like gold. However, Bouri et al.
(2017a) showed that Bitcoin is a poor hedge and is suitable for diversification
purposes only in portfolios composed of stocks, bonds, oil, commodities, and the
USD. On the other hand, Bouri et al. (2018b) found that Bitcoin can act as a safe
haven asset against the global financial stress. On the other hand, Akhtaruzzaman
et al. (2019) found that there is a lower dynamic conditional correlation between
Bitcoin and sectoral stocks and bonds. More importantly, the authors showed that
the Utilities sector has the most effective diversification benefit with Bitcoin.
Group 3 of Table 1, it has been shown that Bitcoin can be a
hedge and a safe haven for traditional assets, but this can be
time varying and depends on the country
(Borri, 2019a; Chan et al., 2013; Katjtazi and Moro, 2019; Shahzad et al., 2019).
Contributions
In this context, our study contributes to the existent literature on Bitcoin and cryptocurrencies in
different ways. First, we consider the specific case of China in which Bitcoin and
cryptocurrencies are forbidden by the government though it is one of the favorite investment
assets of the middle class in China (as mentioned in the Introduction).

Second, we consider the impact of the sector for stocks in China. To the best of our knowledge,
this has not been studied for China with Bitcoin while it has been proven to be important with
gold (e.g., Beckmann et al., 2017; Hoang et al., 2018a).

Third, we model the relationship between Bitcoin, and each considered asset (14 sectoral
stock indices, government bonds and corporate bonds) by simulating a joint distribution
function of returns based on the multivariate Student-t copula. To the best of our knowledge,
this method has not been used to measure the risk of portfolios diversified with Bitcoin. This
method allows us to evaluate the distribution of returns of considered portfolios appropriately
and then to measure the risk of loss on its left tail (with the Value-at-Risk and Expected
Shortfall measures). To this regard, the estimation of the joint distribution using the copula
approach is the main contribution of our study.
Data &
Methodology
Data
The data set of this paper is
composed of 18 time-series
including Bitcoin prices, gold prices
for the Au9995 asset from the
Shanghai Gold Exchange, various
stock indexes and bond indexes
from the Shanghai Stock Exchange
from 20 July 2010 (the first day
Bitcoin quoted a price) to 30 April
2020.

As shown in Table above, the price of all considered assets is

expressed in RMB, the local Chinese currency because we consider
the role of Bitcoin in Chinese portfolios for Chinese investors. However,
for Bitcoin prices, we first consider the USD because it is the reference
currency to express the Bitcoin price worldwide.
Descriptive
it is found that Bitcoin has the highest return in
Statistics the 2010-2020 period (with an average of 115%
per year)

However, its volatility is

very high too, with a
standard deviation of
almost 101% annually.
Three Types of Portfolios
The first one (Type-1) The second one (Type 2) The third one (Type-3)
is composed of only is composed of four assets is also composed of four
three assets which are which are Bitcoin, assets which are gold,
stocks, corporate stocks, corporate bonds, stocks, corporate bonds,
bonds, and government and government bonds. and government bonds.
bonds.

The optimal weight of each asset is determined by the mean-CVaR optimization

method. Since there are 14 different sectoral stock indexes, the total number of
simulated portfolios is 42 portfolios (14 sectors * 3 types).
Methodology
CVaR
Improve the portfolio optimization
procedure by minimizing the
conditional Value at Risk (CVaR)
instead of minimizing the variance
because the distribution of returns
is not normal
Multivariate Student-t
copula
Use the multivariate
student-t copula instead of
bivariate copulas in Ly et
al (2016)
Stochastic Dominance
we also consider the stochastic
dominance method to compare
the distributions of returns of
portfolios with and without Bitcoin
or gold, instead of only distortion
risk like in Ly et al. (2016)

First Second Third Fourth Fifth

Monte Carlo simulation method Distortion Risk

Calculate the integrals by providing provide the procedure to calculate the
new algorithm based on the Monte risk of loss based on the left tail of the
Carlo simulation method portfolio’s return distribution
Result &
Discussions
 Mean CVaR optimal portfolios

As mentioned above, we build optimal portfolios composed of four

assets Bitcoin (or gold), one sectoral stock index, corporate bonds,
and government bonds by minimizing their CVaR. The results are
presented in Table 4
 Mean-CVaR results
Type-1 portfolios
From Table 4, for type-1 portfolios,
Stock C_Bonds T_Bonds
composed of stocks, corporate bonds,
Port. 1: Composite 0.492 0.062 0.436 and government bonds, we see that
Port. 2: Energy 0.49 0.46 0.04
Port. 3: Materials 0.492 0.482 0.016
most of the portfolios are composed of
Port. 4: Industrials 0.492 0.062 0.436 stocks and government bonds with
Port. 5:
Discretionary 0.492 0.482 0.016
about 49% on stocks and 43% on
Port. 6: Staples 0.492 0.482 0.016 government bonds, and a very small
Port. 7: Health care 0.492 0.482 0.016
Port. 8: Financials 0.492 0.062 0.436 part of corporate bonds.
Port. 9: Info. Tech. 0.492 0.482 0.016 This result is almost the same for all
Port. 10: Telecom 0.491 0.008 0.491
Port. 11: Utilities 0.492 0.482 0.016 the stock sectors.
Port. 12:
Commodity 0.492 0.482 0.016
 Mean-CVaR results
Type-2 portfolios

Bitcoin Stock C_Bonds T_Bonds

Port. 1: Composite 0.044 0.464 0.476 0.024 For portfolios of type 2, including also
Port. 2: Energy
Port. 3: Materials
0.042
0.054
0.456
0.458
0.424
0.374
0.068
0.120 Bitcoin, we see that the part of
Port. 4: Industrials 0.076 0.430 0.440 0.046 Bitcoin in optimal portfolios is very
Port. 5: Discretionary 0.056 0.482 0.428 0.034 small, less than 1%.
Port. 6: Staples 0.040 0.469 0.332 0.156
Port. 7: Health care 0.029 0.462 0.072 0.428
This result means that to minimize
Port. 8: Financials 0.040 0.462 0.308 0.182
Port. 9: Info. Tech. 0.044 0.460 0.014 0.486
the CVaR of optimal portfolios,
Port. 10: Telecom 0.076 0.426 0.434 0.058 it should be included a very small
Port. 11: Utilities 0.026 0.474 0.484 0.006
portion of Bitcoin.
Port. 12: Commodity 0.082 0.414 0.014 0.486
Port. 13: A-shares 0.044 0.464 0.476 0.024
Port. 14: B-shares 0.074 0.424 0.486 0.010
 Mean-CVaR results
Type-3 portfolios For portfolios of type 3, show that
Gold Stock C_Bonds T_Bonds
optimal weight of gold in mean-
Port. 1: Composite 0.364 0.148 0.394 0.090 CVaR optimal portfolios is much
Port. 2: Energy 0.396 0.114 0.168 0.330 higher than that of Bitcoin, with a
Port. 3: Materials 0.438 0.068 0.306 0.186
value between 36% and 43%.
Port. 4: Industrials 0.373 0.123 0.148 0.348
Port. 5: Discretionary
0.416 0.086 0.438 0.058 This result means that gold is a better
Port. 6: Staples 0.336 0.164 0.492 0.000 component in Chinese portfolios to minimize the
Port. 7: Health care 0.401 0.090 0.338 0.162 potential loss which is measured by the CVaR.
Port. 8: Financials 0.362 0.142 0.012 0.478 We also note that when gold is included, the
Port. 9: Info. Tech. 0.416 0.080 0.179 0.320 proportion of corporate bonds also becomes
higher than in type-1 portfolios, like with Bitcoin.
Port. 10: Telecom 0.390 0.116 0.096 0.404 This result suggests that corporate bonds are
Port. 11: Utilities 0.354 0.142 0.210 0.286 more suitable when being diversified with
Port. 12: Commodity 0.432 0.082 0.094 0.394 alternative assets like Bitcoin and gold.
Port. 13: A-shares 0.364 0.148 0.394 0.090
To summarize, Table 4 informs us that the optimal weight of Bitcoin in mean-CVaR
portfolios is very low, less than 1%, while the optimal weight of gold in mean-CVaR
portfolios is much higher, between 30% and 40%.

This high weight of gold in Chinese optimal portfolios confirms the result obtained by
Beckmann et al. (2017) who also investigated the role of gold in sectoral stocks in China.

Furthermore, we also note that corporate bonds are more suitable than government bonds
when being diversified with alternative assets like Bitcoin and gold. This result may be
explained by the fact that corporate bonds reflect dynamics of corporations and are thus more
suitable to volatile markets like those of Bitcoin and gold.
 Dependence
structure
with copulas

As explained in Section 3, the multivariate Student-t copula

is used to describe the multivariate dependence structure
among Bitcoin/Gold and traditional assets in China such as
stocks and bonds.

Table 5 presents the results for Bitcoin while Table 6 presents

the results for Gold.
 Parameter 2

Type of (degree of
Copula (Bitcoin USD) copula Parameter 1 (correlation matrix) freedom)

Table 5: Copula Bitcoin + Composite +

C_Bonds + T_Bonds
Student
-t
[[0.0340 -0.0391 -0.0261 0.0559
0.0046 0.3099]] 13.077
dependence Bitcoin + Energy + C_Bonds + Student
T_Bonds -t
[[0.0231 -0.0372 -0.0268 0.0269 -
0.0085 0.3099]] 12.09
structure for Bitcoin + Materials + C_Bonds Student [[0.0402 -0.0377 -0.0255 0.0620 -
+ T_Bonds -t 0.0001 0.3098]] 13.013
portfolios with Bitcoin + Industrials +
Your Text Here
Student [[0.0247 -0.0387 -0.0267 0.0548

Bitcoin C_Bonds + T_Bonds

I hopeBitcoin
and I believe that this +
+ Discretionary
-t
Student
0.0169 0.3096]]
[[0.0360 -0.0395 -0.0257 0.0514
13.222

Template will your Time,

C_Bonds + T_Bonds Money -t 0.0069 0.3115]] 12.334
and Reputation.
Bitcoin + Staples + C_Bonds Student [[0.0343 -0.0386 -0.0268 0.0606
+ T_Bonds -t 0.0110 0.3091]] 14.32
Your Picture Here
Bitcoin + Health care + Student [[0.0261 -0.0407 -0.0252 0.0694
C_Bonds
Your Text Here + T_Bonds -t 0.0185 0.3104]] 14.415

From Table 5, we note that I hopeBitcoin

C_Bonds
Template
+ Financials
and I believe that this+
+ T_Bonds
will your Time, Money
Student
-t
[[0.0079 -0.0408 -0.0271 0.0232 -
0.0153 0.3081]] 12.08
Bitcoin has a positive and low and Reputation.
Bitcoin + Info. Tech. +
C_Bonds + T_Bonds
Student
-t
[[0.0504 -0.0391 -0.0245 0.0551
0.0086 0.3102]] 14.164
correlation with stocks, while it Bitcoin + Telecom + C_Bonds Student [[0.0403 -0.0394 -0.0256 0.0596
+ T_Bonds -t 0.0104 0.3092]] 16.769
has a negative correlation with Your Text Here
Bitcoin + Utilities + C_Bonds Student [[0.0308 -0.0395 -0.0257 0.0701
I hope and I believe that this
corporate and government + T_Bonds
Template will your Time, Money
-t 0.0248 0.3103]] 13.382
Bitcoin + Commodity +
and Reputation. Student [[0.0376 -0.0368 -0.0255 0.0462 -
bonds. C_Bonds + T_Bonds -t 0.0093 0.3099]] 12.44
Bitcoin + A-shares + C_Bonds Student [[0.0338 -0.0391 -0.0261 0.0559
+ T_Bonds -t 0.0047 0.3099]] 13.06
Bitcoin + B-shares + C_Bonds Student [[0.0407 -0.0386 -0.0264 0.0314 -
13.188
+ T_Bonds -t 0.0246 0.3091]]
 Parameter 2

Type of (degree of
Copula (gold) copula Parameter 1 (covariance matrix) freedom)
Table 6: Copula Gold + Composite + C_Bonds +
T_Bonds
Student
-t
[[0.0381 -0.0022 0.0447 0.0543
0.0055 0.3100]] 9.154
dependence Gold + Energy + C_Bonds +
T_Bonds
Student
-t
[[0.0668 -0.0009 0.0454 0.0254 -
0.0082 0.3096]] 9.045
structure for Gold + Materials + C_Bonds + Student [[0.1300 0.0001 0.0468 0.0610 0.0007

portfolios with gold T_Bonds

Your Text Here
Gold + Industrials + C_Bonds +
-t
Student
0.3094]]
[[0.0203 -0.0024 0.0460 0.0523
9.824

I hope and I believe that this

T_Bonds -t 0.0156 0.3098]] 9.09
Template will your Time, Money
and Reputation.
Gold + Discretionary + C_Bonds Student [[0.0269 -0.0026 0.0444 0.0491
+ T_Bonds -t 0.0083 0.3118]] 8.968
Gold + Staples + C_Bonds + Student [[0.0118
Your -0.0029
Picture Here 0.0447 0.0588
T_Bonds -t 0.0132 0.3094]] 10.821
Your Text Here
Gold + Health care + C_Bonds + Student [[0.0039 -0.0029 0.0445 0.0664
Regarding gold, Table 6 shows that the I hope and I believe that this
T_Bonds -t 0.0188 0.3111]] 10.813
correlation between gold and sectoral
Template
Gold will your Time,
+ Financials Money+
+ C_Bonds Student [[0.0225 -0.0022 0.0458 0.0228 -
and Reputation.
T_Bonds -t 0.0140 0.3077]] 8.606
stocks is lower than that with Bitcoin. Gold + Info. Tech. C_Bonds + Student [[0.0100 -0.0001 0.0450 0.0524
T_Bonds -t 0.0088 0.3111]] 10.102
Your +Text
Gold Here + C_Bonds +
Telecom Student [[0.0060 -0.0022 0.0448 0.0600
In some cases, the correlation between T_Bonds
I hope and I believe that this -t 0.0137 0.3108]] 10.824
gold and stocks is even negative, such Template will your Time, Money
Gold + Utilities + C_Bonds + Student [[-0.0031 -0.0011 0.0460 0.0681
and Reputation.
as with stocks of the Utilities and B- T_Bonds -t 0.0262 0.3109]] 9.182

Shares sectors. Gold + Commodity+ C_Bonds +

T_Bonds
Student
-t
[[0.1458 0.0003 0.0457 0.0442 -
0.0080 0.3095]] 9.567
Gold + A-shares + C_Bonds + Student [[0.0383 -0.0022 0.0447 0.0543
T_Bonds -t 0.0055 0.3100]] 9.146
Distortion risk results
Results in Table 7 show that including gold in Chinese portfolios is helpful to reduce
the risk of loss (measured by VaR and ES) and reduce the volatility risk (measured by
the standard deviation). However, including Bitcoin in Chinese portfolios produces the
reverse: Bitcoin does not help reduce the risk but helps increase the return.

This result suggests that gold is more relevant to risk-averse investors whose objective
is to invest in lower-risk portfolios. On the other hand, Bitcoin is more relevant to risk-
seeking investors whose objective is to invest in higher-risk portfolios (Hoang et al.,
2015a). As we mentioned in the Introduction, most of Chinese investors are risk averse.
 Risk 1.
measur Composit Port. 1 – Port. 1 – Port. 1 – Port. 2 – Port. 2 - Port. 2 - 3. Port. 3 – Port. 3 - Port. 3 -
es e S. B G 2. Energy S. B G Materials S B G
VaR 1% -5.731 -2.805 -2.606 -1.612 -6.414 -3.128 -2.910 -1.730 -6.578 -3.220 -3.007 -1.956
VaR 5% -2.324 -1.135 -1.155 -0.619 -2.823 -1.373 -1.352 -0.669 -3.000 -1.466 -1.471 -0.676
ES 1% -5.731 -2.805 -2.606 -1.612 -6.414 -3.128 -2.910 -1.730 -6.578 -3.220 -3.007 -1.956
ES 5% -5.079 -2.480 -2.269 -1.218 -5.268 -2.566 -2.425 -1.309 -5.697 -2.785 -2.587 -1.395
Mean 0.005 0.010 0.033 0.016 -0.033 -0.006 0.014 0.011 -0.002 0.009 0.034 0.016
SD 1.317 0.648 0.681 0.388 1.662 0.815 0.811 0.415 1.729 0.852 0.874 0.427
Sharpe 0.366 1.546 4.845 4.101 -2.007 -0.731 1.748 2.537 -0.126 1.108 3.842 3.668
Risk 5.
measur 4. Port. 4 – Port. 4 - Discretionar Port. 5 – Port. 5 - Port. 5 - Port. 6 – Port. 6 - Port. 6 -
es Industrials S. Port. 4 - B G y S. B G 6. Staples S. B G
VaR 1% -6.557 -3.208 -2.962 -1.623 -6.009 -2.936 -2.923 -1.845 -5.325 -2.602 -2.518 -1.468
VaR 5% -2.699 -1.320 -1.383 -0.632 -2.748 -1.341 -1.435 -0.632 -2.580 -1.259 -1.272 -0.632
ES 1% -6.557 -3.208 -2.962 -1.623 -6.009 -2.936 -2.923 -1.845 -5.325 -2.602 -2.518 -1.468
ES 5% -5.407 -2.642 -2.455 -1.241 -5.295 -2.583 -2.546 -1.276 -4.615 -2.250 -2.177 -1.160
Mean 0.001 0.008 0.045 0.014 0.000 0.010 0.035 0.016 0.044 0.032 0.048 0.023
SD 1.596 0.786 0.851 0.394 1.562 0.769 0.845 0.404 1.515 0.746 0.765 0.397
Sharpe 0.047 1.021 5.333 3.537 -0.030 1.335 4.164 4.057 2.881 4.284 6.308 5.720
Risk
measur 7. Health Port. 7 – Port. 7 - 8. Port. 8 – Port. 8 - Port. 8 - 9. Info. Port. 9 – Port. 9 - Port. 9 -
es care S. Port. 7 - B G Financials S. B G Tech. S. B G
VaR 1% -5.575 -2.726 -2.542 -1.764 -5.771 -2.823 -2.661 -1.671 -6.269 -3.068 -2.900 -1.795
VaR 5% -2.699 -1.318 -1.271 -0.612 -2.374 -1.158 -1.142 -0.627 -3.413 -1.668 -1.626 -0.649
ES 1% -5.575 -2.726 -2.542 -1.764 -5.771 -2.823 -2.661 -1.671 -6.472 -3.190 -3.299 -1.795
ES 5% -4.770 -2.328 -2.181 -1.219 -3.992 -1.940 -1.850 -1.206 -5.453 -2.665 -2.513 -1.290
Mean 0.033 0.027 0.036 0.019 0.019 0.017 0.036 0.015 0.025 0.023 0.039 0.017
SD 1.596 0.786 0.763 0.394 1.524 0.750 0.754 0.398 1.989 0.979 0.969 0.412
 Risk 12.
measures 10. Telecom Port. 10– S. Port.10 - B Port.10 - G 11. Utilities Port. 11 – S. Port.11- B Port.11- G Commodity Port. 12–S Port.12 - B Port.12- G
VaR 1% -6.964 -3.403 -3.142 -1.689 -6.151 -3.009 -2.832 -1.557 -6.295 -3.082 -3.118 -1.925
VaR 5% -3.415 -1.668 -1.631 -0.677 -2.360 -1.152 -1.143 -0.603 -2.938 -1.435 -1.456 -0.690
ES 1% -6.964 -3.403 -3.142 -1.689 -6.151 -3.009 -2.832 -1.557 -6.295 -3.082 -3.118 -1.925
ES 5% -6.139 -2.997 -2.714 -1.323 -5.164 -2.520 -2.387 -1.186 -5.432 -2.656 -2.575 -1.400
Mean 0.019 0.017 0.053 0.016 0.001 0.011 0.023 0.014 -0.018 0.001 0.038 0.013
SD 1.951 0.958 0.979 0.419 1.376 0.679 0.680 0.376 1.691 0.833 0.892 0.433
Sharpe 0.997 1.759 5.456 3.829 0.051 1.599 3.341 3.738 -1.090 0.173 4.205 2.941

Risk
measures 13. A-shares Port.13 – S. Port.13 - B Port.13 - G 14. B-shares Port. 14 – S. Port.14- B Port.14- G C.-Bonds T.-Bonds Bitcoin Gold
VaR 1% -5.730 -2.804 -2.606 -1.611 -7.291 -3.566 -3.135 -1.867 -0.147 -0.289 -37.419 -4.535
VaR 5% -2.325 -1.135 -1.155 -0.619 -2.422 -1.182 -1.281 -0.617 -0.028 -0.028 -8.829 -1.441
ES 1% -5.730 -2.804 -2.606 -1.611 -7.291 -3.566 -3.135 -1.867 -0.147 -0.289 -37.419 -4.535
ES 5% -5.078 -2.479 -2.268 -1.218 -5.820 -2.840 -2.594 -1.331 -0.028 -0.028 -11.424 -3.033
Mean 0.005 0.010 0.033 0.016 -0.001 0.007 0.044 0.015 0.021 0.015 0.461 0.015
SD 1.318 0.649 0.681 0.388 1.430 0.703 0.782 0.389 0.037 0.041 6.434 0.910
Sharpe 0.366 1.546 4.843 4.100 -0.070 1.018 5.652 3.952 56.980 35.541 7.160 1.657
 Stochastic dominance results
In this sub-section, we present the results obtained with the stochastic dominance method.
The objective is to compare the return distributions of the three types of portfolios
considered. The results for the four stochastic dominance orders are presented in Table 8.

Overall, Table 8 shows that for 8
sectors over 14, type-1 portfolios
dominate type-2 portfolios, while
for the 6 other sectors, type-1
portfolios are dominated by type-
2 portfolios. This result means that
in most cases, portfolios without
Bitcoin dominate those with
Bitcoin.
Regarding gold, Table 8 shows that in
all cases, portfolios with gold
dominate portfolios without gold
(type-1 portfolios are dominated by
type-3 portfolios for all sectors).
Regarding the stochastic dominance
comparison between portfolios
including Bitcoin and those including
gold,
 To conclude, the results from the
whole period (2010-2020) show that
gold is a better portfolio diversifier
than Bitcoin because it helps better
reduce the risk. However, Bitcoin
can also be considered as a good
portfolio considering that it helps
increase the return.
From this finding, we conclude that gold is
more suitable to risk-averse investors while
Bitcoin is more suitable to risk-seeking
investors. However, the above results are
valid for the whole period and for daily data.
Then, one can wonder whether these results
still hold if we consider other periods or other
data frequencies
Robustness Tests

3 Robustness
Check
Robustness-check 1: Sub-period analysis

For the first robustness check, we divide the whole period (2010-2020) into two sub-periods. The first one
spreads from 2010 to 2015, during which the price of Bitcoin was low and stable (see Figure 2).
The second one spreads from 2016 to 2020 during which there were high prices and high volatility for
Bitcoin.

First, the weights of both Bitcoin and gold in the optimal portfolios change over time and is higher in sub-
period 2 than in sub-period 1. Second, the dependence structure also changes over time due to the variation
of the value of the parameters of the multivariate Student-t copula. Third, the risk of loss is lower in sub-
period 2 than in sub-period 1. Fourth, the Sharpe ratio is higher for portfolios with Bitcoin than for those
with gold in most of cases only in the first sub-period.

These results thus show the time-varying character of the portfolio diversification between traditional
assets and alternative assets such as Bitcoin and gold. However, this robustness check allows us to confirm
the finding for the whole period because gold remains to be more relevant to risk-averse investors while
Bitcoin remains to be more relevant to risk-seeking investors in both sub-periods 1 and 2.
So, we conclude that this finding is robust.
Robustness-check 2: Monthly analysis
The second robustness check consists of
comparing between daily and monthly results
over the same period from 2010 to 2020.
The objective is to see whether the data
frequency has an impact on the results. This
robustness check also allows us to see whether
the investment horizon (daily or monthly in
this case) can have an impact on the portfolio
diversification using Bitcoin or gold.
The results show that the time
frequency has an impact on the weight
of Bitcoin or gold in optimal portfolios.
It also has impacts on the dependence
structure between the traditional and
alternative assets and on the risk of loss
of the portfolios. However, we still find
that Bitcoin is more relevant to risk-
seeking investors while gold is more
relevant to risk-averse investors.
Robustness-check 3: Does the currency of Bitcoin prices matter?
In this sub-section, we convert Bitcoin
prices into the RMB, while the previous
results are obtained with Bitcoin prices in
the USD, to see whether the currency can
impact our main findings. The results still
show that the finding following which
Bitcoin is more relevant to risk-seeking
investors and gold is more relevant to risk-
averse investors is robust to the change of
the currency of Bitcoin prices (from USD to
RMB).
Finally, the three above robustness
checks show that the results about the
optimal weight of Bitcoin or gold, the
dependence structure and the distortion
risk values can be sensitive to time,
frequency and currency.
However, the conclusion that Bitcoin is
more relevant to risk-seeking investors and
gold is more relevant to risk-averse
investors remains robust.
Conclusion
 Conclusions
• We conclude that gold is more relevant to risk-averse investors (with lower risk and
higher return) while Bitcoin is more relevant to risk-seeking investors (with much
higher risk and much higher return).
• In this context, we conclude that gold is more relevant to Chinese investors. Regarding the
risk, it is important to note that in most cases, the Value at Risk and Expected Shortfall are
about -3% for portfolios with Bitcoin and about -1.5% for portfolios with gold, knowing that
those for non-diversified portfolios are about -3.5%.
• Furthermore, the stochastic dominance results show that portfolios diversified by gold
dominate portfolios diversified by Bitcoin. This result once again demonstrates that gold is
more suitable to risk-averse investors while Bitcoin is more suitable to risk-seeking investors.
Finally, three robustness tests show that this finding is robust to time, data frequency, and the
currency of Bitcoin prices (in USD or in RMB).
 Conclusions
Overall, we conclude that Bitcoin is a better portfolio diversifier in Chinese portfolios for risk
seekers while gold is a better portfolio diversifier for risk-averse investors.

• The prohibition to trade Bitcoin may support the illusion of an even higher potential to the investors
while the result of our research shows that gold is indeed more appropriate to Chinese investors
because most of them are risk-averse (according to the survey of Xiaobo Wu in 2018).

• Though its risky aspect, Bitcoin has been one of the favorite investment assets of the new middle
class in China. May this irrationality be explained by behavioral aspects?

• For future studies, we suggest investigating the relationship between investors’ sentiments,
measured by various proxies such as VIX, Tweets, Google Search, surveys, etc., and the dynamics of
Bitcoin prices, especially during the Covid-19 pandemic.
 References
Aggarwal, D., 2019. Do bitcoins follow a random walk model? Research in Economics, In Press.

Aharon, D.Y., Qadan, M., 2019. Bitcoin and the day-of-the-week effect. Finance Research Letters 73(1), 15-22.

Akyildirim, E., Goncu, A., Sensoy, A. 2020. Prediction of cryptocurrency returns using machine learning. Annals of Operations Research, In Press.

Akyildirim, E., Corbet, S., Katsiampa, P., Kellard, N., Sensoy, A. 2019a. The development of Bitcoin futures: Exploring the interactions between cryptocurrency derivatives.
Finance Research Letters, In Press.

Akyildirim, E., Corbet, S., Lucey, B. M., Sensoy, A., Yarovaya, L. 2019b. The relationship between implied volatility and cryptocurrency returns. Finance Research Letters, In
Press.

Akhtaruzzaman, M., Sensoy, A., Corbet, S. 2019. The influence of Bitcoin on portfolio diversification and design. Finance Research Letters, In Press.

Alaoui, M.E., Bouri, E., Roubaud, D., 2019. Bitcoin price-volume: A multifractal cross-correlation approach. Finance Research Letters, In Press.

Alvarez-Ramirez, J., Rodriguez, E., Ibarra-Valdez, C., 2018. Long-range correlations and asymmetry in the Bitcoin market. Physica A 492, 948-955.

Al-Yahyaee, K.H., Mensi, W., Yoon, S.M., 2018. Efficiency, multifractality, and the long-memory property of the Bitcoin market: A comparative analysis with stock, currency, and
gold markets. Finance Research Letters 27, 228-234.

Ammous, S., 2018. Can cryptocurrencies fulfil the functions of money? The Quarterly Review of Economics and Finance 70, 38-51.

Aslan, A., Sensoy, A. 2019. Intraday efficiency-frequency nexus in the cryptocurrency markets. Finance Research Letters, In Press.

Balcilar, M., Bouri, E., Gupta, R., Roubaud, D., 2017. Can volume predict Bitcoin returns and volatility? A quantiles-based approach. Economic Modelling 64, 74-81.

Bariviera, A.F., Basgall, M.J., Hasperué, W., Naiouf, M., 2017. Some stylized facts of the Bitcoin market. Physica A 484, 82-90.