Abstract
This paper investigates the impact of the launch of Bitcoin futures on intraday volatility using high-frequency data from multiple cryptocurrency exchanges. While the introduction of futures is often associated with the late-2017 market crash, our analysis reveals a dual-phase effect. Using bias-corrected realized volatility and discrete Fourier transform (DFT) analysis, we find that while volatility initially spiked during an adjustment period, the market ultimately reached a state of greater stability than existed prior to the futures launch.
Introduction
In December 2017, the Chicago Board Options Exchange (CBOE) and the Chicago Mercantile Exchange (CME) introduced Bitcoin futures, marking a significant milestone in the institutionalization of cryptocurrency. This event sparked a vigorous debate among economists and market participants. One camp, the "stabilizers," argued that futures would allow informed speculators to enter the market, thereby reducing noise and stabilizing prices. Conversely, "destabilizers" suggested that futures might siphon liquidity from the spot market or facilitate aggressive short-selling, leading to increased volatility.
Most early research focused on the immediate aftermath of the launch—a period characterized by a dramatic price collapse and extreme price swings. This study extends the research horizon to determine if the observed turbulence was a permanent structural change or a temporary transitionary phase as the market adjusted to a new institutional instrument.
Methodology
The study utilizes high-frequency, one-minute log returns from five major exchanges: Bitflyer, Coincheck, Bitstamp, Coinbase, and Binance. The data spans from June 1, 2017, to June 26, 2018, divided into four distinct periods:
- Period 0: Pre-futures (Baseline)
- Period 1: Immediate post-launch (Months 1-2)
- Period 2: Intermediate post-launch (Months 3-4)
- Period 3: Long-term post-launch (Months 5-6)
To ensure robust results, the authors employed several sophisticated statistical tools:
- Bias-Corrected Realized Volatility: This metric accounts for market microstructure noise inherent in high-frequency data to estimate daily volatility accurately.
- Discrete Fourier Transform (DFT): This allows for the decomposition of volatility into low, medium, and high-frequency bands, identifying whether changes were driven by long-term trends or rapid, high-frequency "noise."
- Difference-in-Differences (DD) Analysis: By using Ethereum (which lacked futures at the time) as a control group on the Binance exchange, the authors isolated the specific impact of Bitcoin futures from general market trends.
- Markov-Switching GARCH: A two-regime model used to identify fundamental shifts between high-volatility and low-volatility market behaviors.
Results and Discussion
The Volatility Trajectory
The empirical evidence confirms a distinct "adjustment period." In the pre-launch phase (Period 0), the average realized volatility across exchanges was approximately 0.0502. Immediately following the introduction of futures (Period 1), this figure jumped to 0.0721.
However, this increase was not the "new normal." By Period 3 (six months post-launch), realized volatility plummeted to 0.0295, a level significantly lower than the pre-futures baseline. This suggests that the initial chaos was a transitory phase as the market integrated a new financial instrument.
Frequency Analysis and Regime Shifts
The DFT analysis provided deeper insight, showing that the eventual stabilization occurred across all frequency bands (low, medium, and high). This indicates a holistic maturing of the market rather than a reduction in just one type of price movement, such as flash crashes or long-term price swings.
Furthermore, the Markov-switching model identified a clear regime change. The probability of the Bitcoin market being in a "high-volatility regime" was nearly 100% during the launch period but dropped toward zero by April 2018, signaling a fundamental shift in market behavior.
Causal Evidence via Ethereum
To confirm that this stabilization was due to the introduction of futures and not just a general cooling of the cryptocurrency market, the Difference-in-Differences (DD) analysis compared Bitcoin to Ethereum. The DD estimate for Period 3 ($\beta_3$) was -0.2872 with a t-statistic of -2.6845. This statistically significant result indicates that Bitcoin's volatility reduction was unique and attributable to the introduction of the futures market, rather than a broader market trend.
Conclusion
The introduction of Bitcoin futures initially acted as a destabilizing force, likely due to the sudden influx of new trading strategies and the burst of the 2017 bubble. However, in the long term, the futures market appears to have contributed to price discovery and market maturity. Six months after the launch, Bitcoin exhibited lower intraday volatility than it did before the instruments existed. These findings suggest that while new financial products may cause short-term friction, they can ultimately lead to more stable and efficient underlying markets.
Limitations and Future Work
While the results are robust across multiple exchanges, the study is limited by the specific historical context of the 2017-2018 cycle. Additionally, sample size constraints prevented the use of MCMC simulations for GARCH modeling, necessitating Maximum Likelihood Estimation. Future research should investigate the specific behavior of market makers and liquidity providers during these transition periods to better understand the mechanics of the initial volatility spike.
Summary updated