You are on page 1of 7

Finance Research Letters xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Finance Research Letters


journal homepage: www.elsevier.com/locate/frl

Is the introduction of futures responsible for the crash of Bitcoin?


Liu Ruozhoua,b, , Wan Shanfengc, Zhang Zilib, Zhao Xuejunb

a
Guanghua School of Management, Peking University, Beijing, China
b
Harvest Fund Management Co., Ltd, Beijing, China
c
School of International Trade and Economics, University of International Business and Economics, Beijing, China

ARTICLE INFO ABSTRACT

Keywords: The price of Bitcoin reached its peak a mere few days after the introduction of Bitcoin futures and
Bitcoin suffered an 80% loss in the following year. In this paper, we find a significant and negative
Cryptocurrencies relationship between the introduction of Bitcoin futures and Bitcoin returns, and an insignificant
Bitcoin futures or positive relationship for other 7 major non-Bitcoin cryptocurrencies. Within the first 45 days
Market crash
after the futures launch, Bitcoin suffered a –26.50% loss, while other cryptocurrencies could still
provide positive returns. We presume that the launch of Bitcoin futures was to an extent re-
sponsible for the crash of Bitcoin.

1. Introduction

After a long bull market period, the price of Bitcoin hit its historical peak on December 16, 2017, only a few days after the
introduction of the first Bitcoin futures on December 11, and suffered an 80% loss in the next year. One benefit of futures is that they
make it easy to take a short position. Bitcoin is a highly speculative asset and is sometimes considered one of the biggest bubbles in
history. According to pessimists, the launch of Bitcoin futures made it much easier to burst the bubble.
The price dynamic of Bitcoin is similar with the boom and bust of the home financing market in the 2000s. Fostel and
Geanakoplos (2012) propose the possibility that the crisis might have been caused by financial innovation: tranching and leverage
raised the house price because they attracted optimists and the introduction of CDS lowered the house price because it allowed the
pessimists to bet against the bubble. Naturally, a question arises: is the introduction of futures responsible for the crash of Bitcoin? In
this paper, we try to answer this question by studying the effect of this event on Bitcoin's price and return.
Many studies have considered the relationship between Bitcoin futures and the spot market. Köchling et al. (2018) find that the
efficiency of Bitcoin improved after the futures launch. Corbet et al. (2018) find that the introduction of Bitcoin futures increased the
volatility of Bitcoin. Kapar and Olmo (2019) find that Bitcoin futures market dominates the price discovery process between futures
and spot markets. Sebastião and Godinho (2019) find that Bitcoin futures are effective hedging instruments for both Bitcoin and other
major cryptocurrencies. Recently, Hattori and Ishida (2019) estimate a VAR model to detect whether the actual trade of Bitcoin
futures crashed the Bitcoin market. They find that although futures trading had a negative impact on Bitcoin's price in the first few
minutes, the effect disappeared after 20 min. They conclude that the introduction of Bitcoin futures did not crash the Bitcoin market.
However, as the crash of Bitcoin was a daily or even weekly event, their minutes-level evidence seems too micro to be evidential.
Preliminary studies theoretically suggest that futures markets can attenuate bubbles. For example, Porter and Smith (1995)
studied the asset prices in experimental asset markets. They found that the market bubble can be reduced after introducing a futures
market. Noussair and Tucker (2006) constructed an asset market model in which price bubbles and crashes are widely observed. In


Corresponding author at: Finance Department, Guanghua School of Management, Peking University, Beijing, China.
E-mail address: lrzh1988@163.com (R. Liu).

https://doi.org/10.1016/j.frl.2019.08.007
Received 28 February 2019; Received in revised form 4 August 2019; Accepted 12 August 2019
1544-6123/ © 2019 Elsevier Inc. All rights reserved.

Please cite this article as: Liu Ruozhou, et al., Finance Research Letters, https://doi.org/10.1016/j.frl.2019.08.007
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

Fig. 1. Illustration of study sample.

addition to a spot market, they added a futures market. They found that when futures markets are present, bubbles do not occur in the
spot markets and the futures markets seem to reduce the speculation. Our paper, on the other hand, contributes to the literature by
providing empirical evidence that the introduction of Bitcoin futures was to an extent responsible for the crash of Bitcoin. While
Bitcoin is the only underlying asset of Bitcoin futures, other cryptocurrencies should be immune to the futures launch. Therefore, we
also investigate 7 major non-Bitcoin cryptocurrencies for robustness check. The results are summarized as follows:
Among all the sampled cryptocurrencies, Bitcoin is the first to begin to crash. Shortly after the futures launch, Bitcoin price
crashed the most. Within the first 45 days after the futures launch, Bitcoin suffered a −26.50% loss, while other non-Bitcoin
cryptocurrencies could still provide positive returns.
After controlling for other variables, the daily return of Bitcoin significantly decreased after the futures launch. There exists a
significant and negative relationship between the introduction of Bitcoin futures and Bitcoin return. The average daily return of
Bitcoin decreases −1.172%. For other non-Bitcoin cryptocurrencies, the relationship is either positive or insignificant.
The paper is structured as follows. Section 2 describes the study sample and the sampled cryptocurrencies. Section 3 discusses the
short-term behavior of cryptocurrencies after the futures launch. Section 4 studies the relationship between Bitcoin futures launch
and cryptocurrencies’ returns. Section 5 summarizes and concludes.

2. Study sample and sampled cryptocurrencies

Bitcoin futures first traded on December 11, 2017. We choose a study sample comprising half a year before and after the futures
launch (June 11, 2017 to June 10, 2018). There are 365 observations. The pre-futures period is from June 11, 2017 to December 10,
2017 and the post-futures period is from December 11, 2017 to June 10, 2018. We obtain the US dollar daily close price data of
cryptocurrencies from www.coinmarketcap.com (Fig. 1).
Besides Bitcoin, we also select 7 major non-Bitcoin cryptocurrencies for robustness check. According to www.coinmarketcap.com,
the top 15 Cryptocurrencies (ranked by market cap) in December 10, 20171 are Bitcoin, Ethereum, Bitcoin Cash, IOTA, Ripple,
Litecoin, Dash, Bitcoin Gold, Monero, NEM, Cardano, Ethereum Classic, NEO, Stellar and EOS.
Firstly, we exclude Bitcoin Cash, IOTA, Bitcoin Gold, Monero and EOS because their trading history before Bitcoin futures launch
is less than half a year. On one hand, their price data is incomplete in our study sample. On the other hand, their trading history is too
short to be qualified as a “major cryptocurrencies”. Secondly, within the study sample, we calculate the average daily market cap of
the remaining 10 major cryptocurrencies and rank them. According to Table 1, Bitcoin is the most important cryptocurrency in the
market. The top 7 non-Bitcoin cryptocurrencies are Ethereum, Ripple, Litecoin, Dash, NEM, Stellar and NEO. Finally, we choose
Bitcoin and these 7 non-Bitcoin cryptocurrencies as our sampled cryptocurrencies.

3. Short-term behavior of cryptocurrencies after the futures launch

In this section, we study the short-term behavior of cryptocurrencies. Fig. 2 presents the price dynamics of cryptocurrencies
within 45 days before and after the futures launch (October 27, 2017 to January 24, 2018). After the introduction of Bitcoin futures,
all of the cryptocurrencies continue to rise, but the pace of Bitcoin is already slowing down. After it reaches its peak in the next few
days, Bitcoin is the first to begin to crash. Litecoin and Dash follow its lead, while the other cryptocurrencies continue to boom for
roughly 3–4 weeks.
We also compute the holding period returns of cryptocurrencies in the first 7, 15, 30 and 45 days after the futures launch. The
results are presented in Table 2. Within the first 45 days after the futures launch, Bitcoin suffered a −26.50% loss, while other
cryptocurrencies could still provide positive returns.
To sum up, among all sampled cryptocurrencies, Bitcoin is the first to begin to crash and crashed the most shortly after the futures
launch.

1
https://coinmarketcap.com/historical/20171210.

2
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

Table 1
This table lists the average daily market cap of 10 major cryptocurrencies from June 11, 2017 to June 10, 2018. The data is collected from www.
coinmarketcap.com.
Cryptocurrency Average market cap (millions $) Cryptocurrency Average market cap (millions $)

Bitcoin 127,778 NEM 3505


Ethereum 51,184 Stellar 3492
Ripple 23,468 NEO 3436
Litecoin 6479 Monero 2776
Dash 3601 Ethereum Classic 1997

4. Relationship between Bitcoin futures launch and cryptocurrencies’ returns

Next, we study the relationship between the futures launch and the daily returns of cryptocurrencies by conducting the following
regressions:
Rt = c + 1 Rt 1 + 2 Rt 2 + Futurest + 1 Googlet 1 + 2 Goldt 1 + 3 EPU t 1 + t (1)

Pt is the daily U.S. dollar closing price of cryptocurrencies. We calculate the percentage daily returns as Rt = 100*(Pt − Pt − 1)/
Pt − 1. c is constant and εt is the error term. We build a dummy variable Futurest to represent the Bitcoin futures launch. Futurest is
equal to 0 for the pre-futures period and 1 for the post-futures period. δ measures the change in average daily return after the futures
launch. A negative δ value represents a negative relationship between the futures launch and cryptocurrencies’ returns. If the in-
troduction of Bitcoin futures is to blame for the crash of the underlying asset, then the δ value of Bitcoin is significant and negative,
while those of the 7 non-Bitcoin cryptocurrencies should be either insignificant or positive.
Panagiotidis et al. (2018) considers 21 variables that can affect Bitcoin returns. They find that Google search intensity, gold return
and economics policy uncertainty are the most important variables. Hence, we use them as control variables.
Googlet is the daily global Google Trends search query of the corresponding cryptocurrency. For example, if the dependent
variable is Bitcoin, Googlet is the daily global search frequency of keyword “Bitcoin”. However, when the study sample is larger than 8
months, Google trends only provides weekly data. In order to obtain the daily data, we divide the whole sample into two sub-periods,
collect the daily Google Trends data separately and merge them into one data set. The detailed procedures are presented in appendix.
Goldt is the percentage daily returns of gold. We gather daily gold price data from Bloomberg using the ticker symbol “XAU
Curncy.” EPUt is the percentage daily change of the Economics Policy Uncertainty Index. We use the US Daily News Index, an
improved version of the EPU Index.2 EPUt and Goldt are calculated in the same way as Rt. All of the variables are 7-day/week
frequency daily data with the exception of Goldt. We transform Goldt into 7-day/week frequency by setting it equal to 0 during non-
trading days.
Complicating our consideration is Bitcoin's status as the most important cryptocurrency and the dominant contributor of return
spillovers among major cryptocurrencies (Koutmos, 2018). Even if Bitcoin futures affect only Bitcoin, the crash of Bitcoin should
eventually drag the prices of other cryptocurrencies down. Therefore, regression 1 might cause an inference of spurious relationship.
To address this issue, we employ two augmented regressions as a replacement.
For 7 major non-Bitcoin cryptocurrencies, we conduct regression 2. Rt is the daily% return of non-Bitcoin cryptocurrencies. RtBitcoin
is the contemporary daily% return of Bitcoin.

Rt = c + 1 Rt 1 + 2 Rt 2 + Futurest + 1 Googlet 1 + 2 Goldt 1 + 3 EPU t 1 + RtBitcoin + t (2)

Correspondingly, we conduct regression 3 for Bitcoin. Rt is the daily% return of Bitcoin. Nont is the contemporary market cap
weighted daily% return of all 7 major non-Bitcoin cryptocurrencies.
Rt = c + 1 Rt 1 + 2 Rt 2 + Futurest + 1 Googlet 1 + 2 Goldt 1 + 3 EPU t 1 + Nont + t (3)

Table 3 presents empirical results of regression 2 and regression 3. After controlling for other variables, we find a significant and
negative relationship between the introduction of Bitcoin futures and Bitcoin returns. The average daily return of Bitcoin decreases
−1.172% after the futures launch. For Ethereum, Ripple and Stellar, the relationship is positive. For Litecoin, Dash, NEM and NEO,
even though the sign is negative, the relationship is statistically insignificant. Except NEO, the size of their δ value is much smaller
than that of Bitcoin. In conclusion, there exists a significant and negative relationship between the launch of Bitcoin futures and
Bitcoin return. For other major cryptocurrencies, the relationship is either positive or insignificant.

5. Conclusion

The price of Bitcoin began to crash only a few days after the introduction of Bitcoin futures. Is it responsible for the crash of the
spot market? We try to answer this question by studying and comparing the price and return dynamics of Bitcoin and other cryp-
tocurrencies. As Bitcoin is the only underlying asset of Bitcoin futures, other cryptocurrencies should be immune to the futures

2
www.policyuncertainty.com/us_monthly.html.

3
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

Fig. 2. Price dynamics of cryptocurrencies within 45 days before and after the futures launch.

launch. Therefore, we study 7 major non-Bitcoin cryptocurrencies for robustness checks.


Among all the sampled cryptocurrencies, Bitcoin is the first to begin to crash. Within the first 45 days after the futures launch,
Bitcoin suffered a −26.50% loss, while other non-Bitcoin cryptocurrencies could still provide positive returns. After controlling for
other variables, there exists a significant and negative relationship between the introduction of Bitcoin futures and Bitcoin return. The
average daily return of Bitcoin decreases −1.172% after the futures launch. For other non-Bitcoin cryptocurrencies, the relationship

4
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

Table 2
This table lists the holding period returns of cryptocurrencies in the first 7 days (December 11, 2017 to December 17, 2017), first 15 days (December
11, 2017 to December 25, 2017), first 30 days (December 11, 2017 to January 9, 2018) and first 45 days (December 11, 2017 to January 24, 2018)
after the futures launch.
First 7 days after futures launch First 15 days after futures launch First 30 days after futures launch First 45 days after futures launch

Bitcoin 23.85% −9.24% −5.56% −26.50%


Ethereum 62.99% 73.37% 194.25% 139.69%
Ripple 206.90% 371.91% 784.83% 473.03%
Litecoin 114.40% 83.90% 66.18% 21.07%
Dash 60.59% 70.75% 61.45% 12.55%
NEM 74.59% 156.37% 290.90% 144.63%
Stellar 123.56% 91.60% 391.06% 375.43%
NEO 104.47% 99.10% 281.09% 309.52%

Table 3
This table reports the empirical results of regression 2 and regression 3. For 7 major non-Bitcoin cryptocurrencies, we conduct regression 2. The
empirical results are presented from column 3 to column 9. For Bitcoin, we conduct regression 3. The empirical results are presented in column 2.
The standard errors for coefficients are adjusted following Newey and West (1987) with maximum lag = 5. The symbols *, ⁎⁎, and ⁎⁎⁎ denote
statistical significance at the 10%, 5%, and 1% levels, respectively. T-stats are in parentheses. Futurest is equal to 0 for the pre-futures period and 1
for the post-futures period.
Variables Bitcoin Ethereum Ripple Litecoin Dash NEM Stellar NEO

Rt − 1 −0.003 0.041 0.091 0.086 −0.085 −0.150⁎⁎ 0.066 0.049


(−0.074) (0.843) (1.412) (1.137) (−1.465) (−2.150) (1.017) (0.640)
Rt − 2 −0.020 −0.029 0.045 −0.036 0.004 −0.124⁎⁎⁎ −0.079 0.136
(−0.357) (−0.526) (0.484) (−0.935) (0.138) (−2.913) (−1.500) (1.469)
Futurest −1.172⁎⁎ 0.915* 1.205 −0.027 −0.549 −0.545 0.578 −1.018
(−2.433) (1.819) (1.491) (−0.043) (−0.649) (−0.332) (0.507) (−0.731)
Googlet − 1 −0.043* 0.034⁎⁎ 0.041 0.072 −0.005 0.029 0.017 −0.104
(−1.827) (2.095) (0.913) (1.282) (−0.108) (0.329) (0.249) (−1.077)
Gold t − 1 −0.196 0.569 1.489 −0.513 −0.101 0.016 1.741 0.917
(−0.511) (1.264) (1.627) (−0.613) (−0.207) (0.007) (1.628) (0.773)
EPU t − 1 0.003 −0.003 −0.019⁎⁎⁎ −0.009⁎⁎ −0.002 −0.007 −0.001 0.002
(0.647) (−0.838) (−3.028) (−2.258) (−0.436) (−1.035) (−0.089) (0.250)
RtBitcoin 0.712⁎⁎⁎ 0.621⁎⁎⁎ 0.835⁎⁎⁎ 0.656⁎⁎⁎ 0.681⁎⁎⁎ 0.855⁎⁎⁎ 0.804⁎⁎⁎
(6.350) (4.958) (6.137) (4.890) (2.612) (7.664) (5.655)
Nont 0.514⁎⁎⁎
(8.026)
Constant 1.324⁎⁎⁎ −1.256⁎⁎ −0.400 0.099 0.875 −0.916 0.206 8.283
(2.933) (−2.434) (−0.882) (0.215) (0.309) (−0.190) (0.178) (1.178)
Obs 365 365 365 365 365 365 365 365

is either positive or insignificant. Based on these findings, we presume that the introduction of Bitcoin futures is, at least to an extent,
responsible for the crash of Bitcoin in 2018.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.frl.2019.08.007.

Appendix. The daily Google Trends data formation

In our paper, we need to obtain daily Google Trends data ranging from June 8, 2017 to June 10, 2018. However, when the study
sample is larger than 8 months, Google trends only provides weekly data. To address this problem, we divide the whole sample into
two sub-periods, collect the daily Google Trends data separately and merge them into one data set. The detailed procedures are
presented as follows:

Step1: Date collection.

We collect the daily Google Trends data ranges from June 8, 2017 to December 11, 2017. It is denoted as pre-futures data set.
Then collect the daily Google Trends data ranges from December 11, 2017 to June 10, 2018. It is denoted as post-futures data set.
December 11, 2017 is the overlapping day of both data sets. In the pre-futures data set, the Google Trends value on December 11, 2017
is denoted as Overlappre. In the post-futures data set, the Google Trends value on December 11, 2017 is denoted as Overlappost (Fig. 3).

5
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

Fig. 3. Illustration of Google Trends data collection.

Step2: Date set rescale

According to the algorithm of Google Trends, the Google Trends value on a particular day is sample-dependent. It means the
Google Trends value of a particular day varies when the sample changes. For example, even though both Overlappre and Overlappost
are the Google Trends value on December 11, 2017, they are not necessarily the same because they are from two different samples.
Therefore, before we merge these two data sets, we have to rescale one of them to ensure Overlappre = Overlappost.
If Overlappre > Overlappost, we rescale the pre-futures period data set by multiplying Overlappost/Overlappre. If Overlappre <
Overlappost, we rescale the post-futures period data set by multiplying Overlappre/Overlappost.

Step3: Date merge

After rescaling, we exclude the extra overlapping day data and pool the two data sets together. The new data set is the Googlet

Fig. 4. An example of daily Google Trends data formation.


6
R. Liu, et al. Finance Research Letters xxx (xxxx) xxx–xxx

variable in Section 4.
In Fig. 4, we present a numerical example of the above procedures.

References

Corbet, S., Lucey, B., Peat, M., et al., 2018. Bitcoin futures—what use are they? Econ. Lett. 172, 23–27.
Fostel, A., Geanakoplos, J., 2012. Tranching, CDS, and asset prices: how financial innovation can cause bubbles and crashes. Am. Econ. J.: Macroecon. 4 (1), 190–225.
Hattori T., Ishida R.Did the introduction of Bitcoin futures crash the Bitcoin market at the end of 2017?. Available at SSRN, 2019.
Kapar, B., Olmo, J., 2019. An analysis of price discovery between Bitcoin futures and spot markets. Econ. Lett. 174, 62–64.
Köchling, G., Müller, J., Posch, P.N., 2018. Does the introduction of futures improve the efficiency of Bitcoin? Financ. Res. Lett. forthcoming.
Koutmos, D., 2018. Return and volatility spillovers among cryptocurrencies. Econ. Lett. 173, 122–127.
Newey, W.K., West, K.D., 1987. A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55 (3), 703–708.
Noussair, C., Tucker, S., 2006. Futures markets and bubble formation in experimental asset markets. Pac. Econ. Rev. 11 (2), 167–184.
Panagiotidis, T., Stengos, T., Vravosinos, O., 2018. On the determinants of Bitcoin returns: a lasso approach. Finan. Res. Lett. 27, 235–240.
Porter, D.P., Smith, V.L., 1995. Futures contracting and dividend uncertainty in experimental asset markets. J. Bus. 68 (4), 509–541.
Sebastião, H., Godinho, P., 2019. Bitcoin futures: an effective tool for hedging cryptocurrencies. Financ. Res. Lett. forthcoming.

You might also like