Informational inefficiency on bitcoin futures

Abstract

This paper investigates the dynamics and drivers of informational inefficiency in the Bitcoin futures market. To quantify the adaptive pattern of informational inefficiency, we leverage two groups of statistics which measure long memory and fractal dimension to construct a global-local market inefficiency index. Our findings validate the adaptive market hypothesis, and the global and local inefficiency exhibits different patterns and contributions. Regarding the driving factors of the time-varying inefficiency, our results suggest that trading activity of retailers (hedgers) increases (decreases) informational inefficiency. Compared to hedgers and retailers, the role played by speculators is more likely to be affected by the COVID-19 crisis. Extremely bullish and bearish investor sentiment has more significant impact on the local inefficiency. Arbitrage potential, funding liquidity, and the pandemic exert impacts on the global and local inefficiency differently. No significant evidence is found for market liquidity and policy uncertainty related to cryptocurrency.

KEYWORDS:

• Market inefficiency
• bitcoin futures
• COT reports
• liquidity
• arbitrage

JEL Classifications:

• C22
• G12
• G14

Disclosure statement

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

Notes

1 The CBOE and the CME are the regulated Bitcoin derivatives exchanges. There are several unregulated Bitcoin derivatives exchanges, for example, BitMEX (Alexander et al. 2020) and Deribit (Hoang and Baur 2020). Alexander and Heck (2020) compare the specifications of Bitcoin futures contracts which are traded on two regulated (CME and Bakkt) and five unregulated (BitMEX, Deribit, Huobi, Kraken, and OKEx) derivatives exchanges, and they offer an insightful discussion about these instruments’ roles in price discovery process. Interested readers are referred to their paper for more details.

2 Our study focuses on the largest regulated Bitcoin futures exchange, CME, because of data availability and quality. As CME is a regulated exchange, the quality of its Bitcoin futures data could be guaranteed, and its trader position data are monitored and disclosed by the CFTC.

3 These factors might not cover all the driving forces because it is possible that some latent (unobservable) factors may also affect the inefficiency in the Bitcoin futures market. However, since they are latent, it is infeasible for us to identify and quantify them by using market data. We appreciate one reviewer about reminding us of this potential limitation of our study.

4 The concept of entropy is borrowed from classical mechanics and information theory, which is normally used to quantify the complexity of a system (Yentes et al., 2012). Several researchers utilise entropy-based measure to quantify information content or randomness of a financial time series, e.g. Benedetto, Giunta, and Mastroeni (2016) and Alvarez-Ramirez, Rodriguez, and Alvarez (2012), among others. An unpredictable (predictable) series has a high (low) entropy value. Pincus’s (1991) approximate entropy method is used in Kristoufek and Vosvrda’s (2014b) study because it is conceptually simple and suitable for a finite time series. Singular value decomposition (SVD) entropy has also been applied in testing stock and cryptocurrency market efficiency, e.g. Caraiani (2014), Gu, Xiong, and Li (2015), and Alvarez-Ramirez and Rodriguez (2021a, 2021b). We do not incorporate entropy-based measures in our efficiency index. The key reason is that according to the definition of entropy, it is difficult to judge whether it measures global efficiency or local efficiency.

5 On 30 August 2019, Gemini was added by the CME as the fifth spot exchange to calculate the Bitcoin futures reference rate (Aleti and Mizrach 2021). On 31 March 2022, LMAX Digital was added by the CME as the sixth spot exchange to calculate the Bitcoin futures reference rate (CFBenchmarks 2022).

6 These two exchanges are selected because they satisfy all the three conditions: (1) The exchange is one of the constituent exchanges used by the CME to calculate the BRR; (2) The exchange contributes data to the BRR over the whole sample period of our study; 3) The exchange’s spot price data during the sample period of our study are available at bitcoincharts.com. Alexander and Dakos (2020) highlight the importance of using reliable data sources to conduct the cryptocurrency research and highlight that some data sources, e.g. CoinMarketCap, may have the issues such as artificial volume and inconsistent time stamps. They suggest that bitcoincharts.com is a reliable data source.

7 More details about the CFTC and the COT reports could be found at: https://www.cftc.gov/MarketReports/CommitmentsofTraders/ExplanatoryNotes/index.htm. Several limitations of the COT data are mentioned by Tornell and Yuan (2012). For example, contract maturities are not considered when aggregating the position data. Also, there is a two-day gap between the collection date (Tuesday) and the release date (the following Friday) of the COT data.

8 Regarding more details about global and local characteristics of a time series, interested readers can refer to Kristoufek and Vosvrda (2013, 2014a, 2014b) and Salcedo-Sanz et al. (2022).

9 To incorporate the weekly COT data and Lucey et al.’s (2022) weekly cryptocurrency policy uncertainty index data in our analysis, the frequency of market inefficiency measures in our study is weekly. However, we would certainly consider using higher frequency data in our follow-up studies if higher frequency COT data and cryptocurrency policy uncertainty index data become available.

10 Baker and Wurgler (2006, 2007) offer excellent surveys of sentiment measures.

11 Please refer to Almost $6B Crypto Futures Liquidated As BTC Falls Below $48K – February 23% (bitcoinprice.com) for more details about this news.