Quantile causal relationship between Bitcoin and stock indices

Abstract

This study employs a Granger non-causality test in quantiles to analyze the causal relationship between Bitcoin and representative stock indices. We further bifurcate our analysis into pre- and post-COVID-19 periods, providing a unique perspective on hedge evaluation in different market conditions. The empirical findings reveal several key insights. First, a traditional causal test conducted over the entire period, which only considers causality at the mean, leads us to reject the null hypothesis that Bitcoin does not Granger cause any of the nine stock indices. However, we find that Bitcoin is not Granger caused by five out of nine stock indices. Second, by extending the analysis to the overall quantile interval, we find significant results in 12 out of 18 cases. Third, we identify robust causal relationships between Bitcoin and stock indices across lower and higher quantile intervals. Lastly, in the post-COVID-19 period, characterized by heightened price volatility and increased uncertainty, we observe a near-universal reversal in the causal relationships between Bitcoin and stock indices. Furthermore, the number of cases exhibiting causality increased markedly compared with the pre-COVID-19 period, which was characterized by more moderate price volatility and uncertainty.

Keywords:

• Bitcoin
• stock indices
• quantile regression
• quantile non-causality test
• COVID-19

JEL CLASSIFICATIONS:

• C22
• G13
• F39

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The Bitcoin data are available at https://coinmarketcap.com

2 The selection of May 1, 2013, as the starting point instead of Bitcoin’s launch date, aims to exclude earlier data with relatively low values and trading volumes. November 1, 2019, was chosen to allow for some periods before the sharp price declines due to the COVID-19 pandemic. Although December 2019, when COVID-19 was first identified, was considered, no significant difference in estimation results was found compared to starting from November 1, 2019. Therefore, November 1, 2019, was selected to account for events preceding the identification of COVID-19. The estimated results can be provided upon request.

3 The time series plots are not reported. They can be provided upon request.

4 The time series figure of the uncertainty index and the results of Granger non-causality tests for the subintervals [0.05,0.95], [0.05,0.5], and [0.5,0.95] of the pre- and post-COVID-19 periods are not reported. They can be provided upon request.

5 I would like to thank an anonymous referee for providing insightful comments and suggestions that significantly improved the quality of this paper.

Additional information

Funding

This research was supported by the Chung-Ang University Research Grants in 2022.