Economic policy uncertainty and cryptocurrency volatility

Economic policy uncertainty and cryptocurrency volatility

Journal Pre-proof Economic Policy Uncertainty and Cryptocurrency Volatility Kuang-Chieh Yen , Hui-Pei Cheng PII: DOI: Reference: S1544-6123(19)31018...

727KB Sizes 1 Downloads 96 Views

Journal Pre-proof

Economic Policy Uncertainty and Cryptocurrency Volatility Kuang-Chieh Yen , Hui-Pei Cheng PII: DOI: Reference:

S1544-6123(19)31018-9 https://doi.org/10.1016/j.frl.2020.101428 FRL 101428

To appear in:

Finance Research Letters

Received date: Revised date: Accepted date:

19 September 2019 18 November 2019 8 January 2020

Please cite this article as: Kuang-Chieh Yen , Hui-Pei Cheng , Economic Policy Uncertainty and Cryptocurrency Volatility, Finance Research Letters (2020), doi: https://doi.org/10.1016/j.frl.2020.101428

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier Inc.

Highlights 

This paper investigates whether the economic policy uncertainty (EPU) index provided by Baker et al. (2016) can predict cryptocurrency volatility.



We show that the China EPU index can predict negatively the Bitcoin monthly volatility.



China’s ban on crypto-trading does not result in a response in the volatility of Bitcoin.

1

Economic Policy Uncertainty and Cryptocurrency Volatility Kuang-Chieh Yen and Hui-Pei Cheng * November 2019

Abstract We investigate the relationship between the economic policy uncertainty index (EPU) and cryptocurrency volatility. We find that a change in EPU of China predicts cryptocurrency volatility, but a change in the EPU of the U.S., Japan, or Korea has no such effect. Moreover, changes in the China EPU are negatively associated with Bitcoin and Litecoin future volatility, which may imply that Bitcoin and Litecoin are hedging tools against the EPU risk. However, changes in China EPU may not affect the cryptocurrency volatility after the Chinese government’s regulation of cryptotrading. JEL: C22, G15, D81 Keywords: Bitcoin, Cryptocurrencies, Economic policy uncertainty, China, Volatility



Hui-Pei Cheng ([email protected]) and Kuang-Chieh Yen ([email protected]) are both Assistant Professors in Department of Economics at Soochow University, Taiwan. Hui-Pei Cheng is the corresponding author. The authors are grateful to the Ministry of Science and Technology of Taiwan for the financial support provided for this study

2

1. Introduction Recent research has explored the predictors of the return on cryptocurrency, since cryptocurrency is increasingly popular with investors because it can act as an alternative financial asset in financial markets. Predictors could be classified into two types: market-based and macro-based factors. Market-based factors include technical indicators (Gerritsen et al., 2019), the three factors of the crypto-pricing model (Shen et al., 2019), and volatility indexes in stock markets (Bouri et al., 2017). Macroeconomic factors include global economic activity (Cheng and Yen, 2019a) and economic policy uncertainty (Demir et al., 2018; Cheng and Yen, 2019b). The economic policy uncertainty index (EPU) of Baker et al. (2016) is considered an important factor in the cryptocurrency market. Demir et al. (2018) observe that uncertainty regarding government decisions can make investors lose trust in their fiat currencies or worry about the overall economy, in particular after the 2008 financial crisis. Thus, a change in the EPU may cause investors to reconsider their portfolio to avoid potential wealth loss. Although the above studies have explored whether the EPU is related to Bitcoin returns, no studies have compared the predictive power of national EPU across countries on cryptocurrency volatility. Therefore, this paper investigates which country’s EPU may play an important role in predicting cryptocurrency volatility. In this paper, we develop two hypotheses based on investor behaviours. The first hypothesis is based on the fear of the investors in the crypto-market. An increase in the EPU will make investors feel that the market conditions of the cryptocurrency are worsening. They may thus move their money from the cryptocurrency market to other financial markets if they think that cryptocurrency returns are negatively related to the

3

economic policy uncertainty. 1 This cash outflow will make the crypto-market less liquid and exaggerate the volatility of the cryptocurrency in the future. The second hypothesis emphasises the hedging effect on the volatility of the cryptocurrency. When the economic uncertainty increases, investors may invest their money in a cryptocurrency if they consider the cryptocurrency to be a safe haven asset. Therefore, the cash inflow to the cryptocurrency market will make the crypto-market more liquid and reduce the volatility of the cryptocurrency in the future. Practically, to avoid the effects of the persistent and mean-reverting property of the volatility, we modify the stochastic volatility model from Wang and Yen (2019) to examine whether national EPU can drive cryptocurrency volatility and test our two hypotheses. We include the EPU of the United States (U.S.), China, Japan, and (South) Korea.2 Our main findings show that only a change in the China EPU is significantly and negatively associated with the future volatility of Bitcoin and Litecoin. This may occur because China possesses the largest mining pools in the world (Ma, et al., 2018). In addition, we do not detect a clear effect of a change in China EPU on the volatility of the Bitcoin after the China government’s crypto-ban policy after September 2017. Overall, our empirical results support that the hedging hypothesis. Therefore, this paper suggests that the cryptocurrency could be a hedge asset against EPU risk. This paper contributes to the literature on volatility forecasting of cryptocurrency in two ways. First, we discuss volatility forecasting using the effect of national EPU on the cryptocurrency market and identify which country is relevant to cryptocurrency volatility. Second, based on our empirical results, we show that cryptocurrency can be a good hedging tool against EPU risk. 1

See the detail discussion in Demir et al. (2018). The four countries are important to the development of the crypto-market. For instance, the U.S. has the largest trading capacity of the Bitcoin futures. Both Japan and Korea, relative to other countries, implement complete regulations for crypto-trade. Furthermore, China has the largest Bitcoin mining pools in the world. 2

4

The remainder of this paper is structured as follows. Section 2 discusses the data, hypothesis and the empirical approach while Section 3 discusses the estimation results. Section 4 concludes.

2. Data, Hypothesis and Empirical Approach 2.1 Data We obtain the daily price of Bitcoin (BTC), Litecoin (LTC), and Ripple (XRP), from the coinmarketcap website. 3 The monthly volatility is calculated as the daily returns in each month. The EPU of China, the United States, Japan, and Korea as constructed by

Baker,

et

al.

(2016)

are

obtained

from

the

website

(http://www.policyuncertainty.com). Based on our available database, the sample period runs from February 2014 to June 2019.

2.2 Hypothesis Development In this section, we develop two hypotheses that discuss how EPU is related to cryptocurrency future volatility. Hypothesis 1:

The fear hypothesis in the crypto-market: EPU will have positive predictive power for cryptocurrency volatility. Investors may short their cryptocurrency position and long other financial assets due to their worries about the current situation in the cryptomarket. The cash outflow will make the crypto-market less illiquid and exaggerate the future volatility of the cryptocurrency.

Hypothesis 2:

Hedging hypothesis in the crypto-market: EPU will have negative predictive power for cryptocurrency volatility. Investors may invest more in the crypto-market if they consider the cryptocurrency to be a safe haven asset. Thus, the cash inflow will make the crypto-market more liquid and decrease the volatility of the cryptocurrency.

3

See for more detail in the website (https://coinmarketcap.com/)

5

2.3 Empirical Approach To develop our empirical strategy, we modify the stochastic volatility model of Wang and Yen (2019) as follows: ( where

)



is the variance mean-reverting speed;

volatility of the volatility;

,

(1)

is the mean of variance;

is the Brownian innovations; and

is the

is the stochastic

variance jump term. Without the loss of generality, we ignore the stochastic variance jump term (

) to simplify the model complexity. We then discretize the Eq. (1) as

follows. (

)

(

The first difference term, Moreover, the term,

)

√ (

)

(2)

, control for the variance persistent property. , controls for the mean-reverting property of the variance.

We expect that the corresponding coefficient ( ) will be positive. The last term, √ (

) , may be considered an error term that follows the normal

distribution with zero mean. Based on the Eq. (2), we form our regression model: ( where

)



(3)

is the specific cryptocurrency volatility which is measured by the monthly

variance that is calculated by the daily returns at time ; and period;

;

is the difference between

is the specific cryptocurrency’s average variance during our sample is the change rate of the EPU of different country including China,

U.S., Japan, and Korea at time t; is the error term at time t.

is the vector of year and quarter dummies; and

is corresponding to the coefficient

thus it is expected to be positive.

, and

6

in the Eq. (2) and

measure the relationship between the

change in EPU of different country and the crypto-volatility. We adopt the NeweyWest standard errors to eliminate the potential heterogeneity and autocorrelation problems (Newey and West, 1987).

3. Empirical Results 3.1 The Effect of EPU on Bitcoin Volatility In this section, we apply Eq. (3) to investigate whether the EPU of these four countries (China, U.S., Japan, and Korea) can predict the Bitcoin monthly volatility. In Table 1, the coefficient

is significantly positive in all Models, which is

consistent with our theoretical model. The coefficients

and

in Model (1) are

both apparently negative, revealing that a higher EPU in China leads to higher Bitcoin volatility in the next two and three months. Furthermore, the coefficients of

are

significant in Model (2) and (4) while the adjusted R-square is weaker than in Model (1). We conjecture that a higher EPU induces investors to exchange their fiat money for Bitcoin, and then increases the liquidity of Bitcoin, reducing Bitcoin volatility. Overall, our empirical results support our Hedging Hypothesis (Hypothesis 2). Since we observe that the EPU of the U.S. and Korea affect the Bitcoin volatility, we go further step to investigate whether the impacts of U.S., Japan, or Korea comes from that of China. We combine the EPU of China and the other countries (U.S., Japan, and Korea) and run the following regression model. (

where

)





, (4)

is the Bitcoin volatility defined as the variance calculated by the daily

returns at time ;

is the change rate of the China EPU at time t;

is the change rate of the U.S., Japan, or Korea EPU at time t in Model (1),

7

(2), and (3), respectively;

is the vector of year and quarter dummies; and

is the

innovation at time t.
In Table 2, the coefficients of those lagged terms of China (

and

) are

both significantly negative in all models while none of the coefficients of the U.S., Japan, and Korea (

,

, and

in all Models) are significant, implying that the

China EPU has the strongest influence on Bitcoin volatility among the EPU of the countries studied. Thus, we find that the EPU of China plays a vital role in Bitcoin volatility, even after controlling for other countries’ EPU. Our findings still support the Hedging Hypothesis (Hypothesis 2).

3.2 The Effect of EPU on Other Cryptocurrency Volatility We have shown that the EPU of China can predict the Bitcoin volatility, while the EPU of the other countries cannot. It is natural to ask whether this phenomenon exists in other cryptocurrencies, for example, Ripple and Litecoin. We thus run the regression model of Eq. (4) and replace the dependent variables with the Litecoin and Ripple volatility. The estimation results are shown in Tables 3 and 4, respectively.
In Table 3, we find that the coefficients (

,

, and

) of China EPU are all

negative and the findings are similar to the Bitcoin case. Moreover, these coefficients are more apparent in Model (1) and (2).
However, the results in Table 4 indicate that all the coefficients

and

are insignificant. The findings suggest that China’s EPU may not affect Ripple volatility. This could be that the payment system of Ripple is more decentralized than that of Bitcoin (Amknecht, et al., 2015). Overall, the findings here still suggest that

8

higher EPU leads to lower cryptocurrency volatility. Thus, these findings support our Hedging Hypothesis.

3.3 Robustness 3.3.1 The Effects of All Countries’ EPUs on Cryptocurrency Volatility In this subsection, we include the EPU of all study countries to determine whether the coefficient of the change rate in the EPU of China remains significant for the cryptocurrency volatility. We run the following regression model, ( where

)





,

(5)

is the specified cryptocurrency volatility defined as the variance calculated

by daily return at time ; C denotes the country set, *China, US, Japan, Korea+, and other variables are defined as above.
In Table 5, the coefficient of the China EPU lagged term,

,

are negative in all Models. Furthermore, we find that only

and is

significant in the case of Bitcoin and Litecoin. We observe that when considering the EPUs of all study countries in the model, the significance of the China EPU weakens. 4 However, these results remain consistent with our main results. Overall, our results still support the Hedging Hypothesis in cases of Bitcoin and Litecoin.

3.3.2 The Regulation of Chinese Crypto-trading Finally, we study whether a change in Chinese government policy affects Bitcoin volatility. We run the following regression model, (

)





, (5)

4

The less significant results may be caused by the small sample size.

9

where all variables are defined as above except for

being the corresponding

dummy that represents the time after the government of China banned the trading of cryptocurrencies in September, 2017.
In Table 6, the coefficients,

and

, are negative but not significant. This

implies that the announcement of the ban on crypto-trading by the Chinese government may not reduce Bitcoin volatility. Hence, this policy change has no significant impact on Bitcoin volatility forecasting using EPU. Overall, this China policy change does not trigger a reaction in the volatility of Bitcoin.

4. Conclusion In this paper, we find that a negative linkage between EPU and Bitcoin future volatility, meaning that higher EPU leads to lower Bitcoin volatility. Furthermore, we determined that the China EPU, rather than that of the other countries, plays an important role in the volatility of cryptocurrencies such as Bitcoin. We obtain similar results for Litecoin. Finally, we suggest that the announcement of the ban on cryptotrading by the Chinese government has no significant impact on the predictive power of the China EPU for the volatility of Bitcoin. Overall, we suggest that the cryptocurrency can be a hedging tool against EPU risk. author_statement Kuang-Chieh Yen and Hui-Pei Cheng: Conceptualization, Methodology, Software, Data curation, Writing-Original draft preparation, Writing-Reviewing and Editing.

10

References Armknecht, F., Karame, G. O., Mandal, A., Youssef, F., & Zenner, E. (2015, August). Ripple: Overview and outlook. In International Conference on Trust and Trustworthy Computing (pp. 163-180). Springer, Cham. Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring economic policy uncertainty. The Quarterly Journal of Economics, 131(4), 1593-1636. Bouri, E., Gupta, R., Tiwari, A. K., & Roubaud, D. (2017). Does Bitcoin hedge global uncertainty? Evidence from wavelet-based quantile-in-quantile regressions. Finance Research Letters, 23, 87-95. Cheng, H. P., & Yen, K. C. (2019a). The relationship between the real economy and the cryptocurrency market. Centre for Business Research and Development, Working paper. Cheng, H. P., & Yen, K. C. (2019b). The relationship between the economic policy uncertainty and the cryptocurrency market. Finance Research Letters, forthcoming. Demir, E., Gozgor, G., Lau, C. K. M., & Vigne, S. A. (2018). Does economic policy uncertainty predict the Bitcoin returns? An empirical investigation. Finance Research Letters, 26, 145-149. Gerritsen, D. F., Bouri, E., Ramezanifar, E., & Roubaud, D. (2019). The profitability of technical trading rules in the Bitcoin market. Finance Research Letters. Forthcoming. Ma, J., Gans, J. S., & Tourky, R. (2018). Market structure in bitcoin mining (No. w24242). National Bureau of Economic Research. Newey, W. K., & K. D. West (1987) A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703–708.

11

Shen, D., Urquhart, A., & Wang, P. (2019). Three-factor Pricing Model for Cryptocurrencies. Finance Research Letters. Forthcoming. Wang, Y. H., & Yen, K. C. (2019). The information content of the implied volatility term structure on future returns. European Financial Management, 25(2), 380-406.

Table 1 Economic policy uncertainty index on Bitcoin volatility This table reports the estimation of the following regression model, ( ) ∑ , where is the Bitcoin volatility defined as the variance calculated by daily returns at time ; is the difference between and ; is the change rate in the economic policy uncertain index of China, U.S., Japan, or Korea at time t in Model (1), (2), (3), and (4), respectively; is the long-run level of the Bitcoin variance during our sample period; is the vector of year and quarter dummies; and is the innovation at time t. We use the NeweyWest standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE Observation

(1) China

(2) U.S.

(3) Japan

(4) Korea

0.897*** (0.156) -0.001 (0.000) -0.001** (0.001) -0.001*** (0.000) 0.000 (0.001)

0.831*** (0.166) -0.000 (0.000) -0.000 (0.000) -0.001* (0.000) 0.000 (0.001)

0.809*** (0.162) -0.001 (0.001) 0.000 (0.001) -0.001 (0.001) -0.000 (0.001)

0.832*** (0.176) -0.000 (0.000) -0.000 (0.000) -0.001* (0.000) 0.000 (0.001)

0.4063 YES YES 63

0.3369 YES YES 63

0.3294 YES YES 63

0.3331 YES YES 63

12

Table 2 Comparisons between China and other countries EPU This table reports the estimation of the following regression model, ∑ ( ) ∑ where is the Bitcoin volatility defined as the variance calculated by daily returns at time ; is the difference between and ; is the change rate of the China economic policy uncertainty index at time t; is the change rate in the U.S., Japan, and Korea economic policy uncertainty index at time t in Model (1), (2), and (3), respectively. is the long-run level of the Bitcoin variance during our sample period; is the vector of year and quarter dummies; and is the innovation at time t. We use the Newey-West standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE Observation

(1) China / U.S.

(2) China / Japan

(3) China / Korea

0.883*** (0.166) -0.001 (0.001) -0.001** (0.001) -0.001** (0.000) -0.000 (0.000) 0.000 (0.000) -0.000 (0.000) 0.000 (0.001)

0.876*** (0.160) -0.001 (0.001) -0.001** (0.001) -0.001** (0.000) -0.001 (0.001) 0.001 (0.001) 0.001 (0.001) 0.000 (0.001)

0.860*** (0.171) -0.001 (0.001) -0.001** (0.001) -0.001** (0.001) -0.000 (0.000) 0.001 (0.000) 0.000 (0.000) 0.000 (0.001)

0.3727 YES YES 63

0.3936 YES YES 63

0.3840 YES YES 63

13

Table 3 The influence of the China EPU on Litecoin volatility This table reports the estimation of the following regression model, ∑ ( ) ∑ where is the Litecoin volatility defined as the variance calculated by daily returns at time ; is the difference between and ; is the change rate in the China economic policy uncertainty index at time t; is the change rate in the U.S., Japan, and Korea economic policy uncertainty index at time t in Model (1), (2), and (3), respectively. is the long-run level of the Bitcoin variance during our sample period; is the vector of year and quarter dummies; and is the innovation at time t. We use the Newey-West standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE Observation

(1) China / U.S.

(2) China / Japan

(3) China / Korea

1.011*** (0.116) -0.003* (0.002) -0.004*** (0.001) -0.002*** (0.001) 0.001 (0.001) 0.001 (0.001) -0.001 (0.002) 0.001 (0.001)

1.013*** (0.112) -0.003* (0.001) -0.005*** (0.002) -0.003*** (0.001) 0.000 (0.002) 0.003 (0.003) 0.002 (0.002) 0.001 (0.001)

1.036*** (0.112) -0.002 (0.002) -0.003* (0.002) -0.002 (0.001) 0.000 (0.001) -0.000 (0.001) -0.001 (0.002) 0.001 (0.001)

0.4927 YES YES 63

0.4893 YES YES 63

0.4846 YES YES 63

14

Table 4 The influence of the China EPU on Ripple volatility This table reports the estimation of the following regression model, ∑ ( ) ∑ where is the Ripple volatility defined as the variance calculated by daily returns at time ; is the difference between and ; is the change rate in the China economic policy uncertainty index at time t; is the change rate in the U.S., Japan, and Korea economic policy uncertainty index at time t in Model (1), (2), and (3), respectively. is the long-run level of the Bitcoin variance during our sample period; is the vector of year and quarter dummies; and is the innovation at time t. We use the Newey-West standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE Observation

(1) China / U.S.

(2) China / Japan

(3) China / Korea

0.918*** (0.119) -0.004 (0.003) -0.007 (0.007) -0.004 (0.004) -0.001 (0.002) 0.002 (0.004) 0.001 (0.005) -0.002 (0.003)

0.918*** (0.116) -0.005 (0.003) -0.006 (0.006) -0.005 (0.004) 0.005 (0.008) -0.003 (0.007) 0.010 (0.006) -0.001 (0.003)

0.919*** (0.121) -0.003 (0.003) -0.006 (0.006) -0.004 (0.004) -0.003 (0.002) -0.001 (0.004) 0.002 (0.005) -0.002 (0.002)

0.3186 YES YES 63

0.3262 YES YES 63

0.3227 YES YES 63

15

Table 5 The influence of the All countries’ EPU on the cryptocurrency volatility This table reports the estimation of the following regression model, ( ) ∑ ∑ , (5) where is the Bitcoin, Litecoin, and Ripple volatility defined as the variance calculated by daily return at time in Model (1), (2), and (3), respectively; is the difference between and ; C denotes the country set, *China, US, Japan, Korea+; is the long-run level of the Bitcoin variance during our sample period; is the vector of year and quarter dummies; and is the innovation at time t. We use the Newey-West standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE All Other Countries Observation

(1) BTC

(2) LTC

(3) XRP

0.833*** (0.186) -0.001 (0.001) -0.002* (0.001) -0.001* (0.001) 0.000 (0.001)

1.010*** (0.100) -0.003 (0.002) -0.004** (0.002) -0.003 (0.002) 0.001 (0.001)

0.924*** (0.136) -0.006 (0.004) -0.007 (0.007) -0.007 (0.005) -0.002 (0.003)

0.3400 YES YES YES 63

0.4439 YES YES YES 63

0.2428 YES YES YES 63

16

Table 6 The impact of the China banned crypto-trading for Bitcoin volatility This table reports the estimation of the following regression model, ( ) ∑ ∑ , where is the Bitcoin volatility defined as the variance calculated by daily return at time ; is the difference between and ; is the change rate of the China economic policy uncertain index at time t; is the change rate of the U.S., Japan, and Korea economic policy uncertainty index at time t in Model (1), (2), and (3), respectively; is the vector of year and quarter dummies; and is the innovation at time t. denotes the corresponding dummy that represents the China government banned the trading of the cryptocurrencies in September 2017. We use the Newey-West standard errors with the 3 lags. The sample period runs from February 2014 to June 2019. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Model

Constant

adj. Year FE Quarter FE All other countries Observation

(1) China / U.S.

(2) China / Japan

(3) China / Korea

(4) All Countries

0.981*** (0.130) -0.001 (0.001) -0.001 (0.001) 0.003* (0.001) -0.001 (0.001) -0.001 (0.001) 0.001 (0.001)

0.959*** (0.126) -0.001* (0.001) -0.001* (0.001) 0.002** (0.001) -0.001 (0.001) -0.001 (0.001) 0.001 (0.001)

0.945*** (0.127) -0.001* (0.001) -0.001 (0.001) 0.003** (0.001) -0.001 (0.001) -0.001 (0.001) 0.001 (0.001)

0.953*** (0.132) -0.001 (0.001) -0.001 (0.001) 0.003* (0.001) -0.002 (0.001) -0.002 (0.001) 0.001 (0.001)

0.4405 YES YES YES 63

0.4592 YES YES YES 63

0.4565 YES YES YES 63

0.4520 YES YES YES 63

17