When Does Crypto Diversify? A Regime Analysis of Bitcoin-Equity Correlation (2013–2025)

In 2022, Bitcoin fell 65% while the S&P 500 fell 18%. If you held both to diversify, you held two versions of the same trade. The year that was supposed to vindicate Bitcoin as an inflation hedge and uncorrelated store of value produced instead its worst-ever correlation with equities — a 30-day BTC-SPX rolling correlation that peaked at approximately 0.71 during the rate-hike cycle.
Contrast that with the 2020 COVID recovery. From April through December of that year, Bitcoin returned roughly 303% while the S&P 500 returned 65% and 10-year Treasuries barely moved. BTC-SPX rolling correlation during that period averaged approximately 0.15. The same asset that amplified losses in 2022 acted as a genuine diversifier in 2020.
The question is not whether Bitcoin diversifies. The historical record shows it sometimes does, and sometimes dramatically fails to. The question is under which conditions — and whether those conditions are systematically predictable.
This article presents Quant Decoded's original backtest spanning 2013 through 2025, using a four-regime framework defined by VIX level and S&P 500 direction to systematically examine when BTC-SPX correlation is low enough to benefit a diversified portfolio and when it is high enough to amplify drawdowns. It then evaluates a simple correlation-monitoring overlay that historically improved risk-adjusted outcomes without sacrificing meaningful upside.
The Unstable Correlation: A Decade of Rolling Data
The conventional narrative — "Bitcoin is uncorrelated with equities" — was largely accurate through 2019. During the 2014–2019 period, the average annual 30-day rolling BTC-SPX correlation rarely exceeded 0.20, and for extended stretches hovered near zero. Bitcoin was then a small, retail-dominated asset with limited institutional participation, whose price movements reflected idiosyncratic factors: mining difficulty adjustments, exchange failures, regulatory announcements specific to crypto markets.
That changed progressively from 2020 onward. The entry of institutional investors — through futures markets, publicly traded holding companies, and eventually spot ETFs approved in the US in January 2024 — integrated Bitcoin into the same risk-on/risk-off framework that governs equities. As Bitcoin became a recognized risk asset, its correlation with the S&P 500 during macro stress events rose substantially.
| Year | Avg 30-Day BTC-SPX Rolling Correlation | Dominant Regime |
|---|---|---|
| 2014 | ~0.05 | Mixed |
| 2015 | ~0.03 | Risk-Off / Low Vol |
| 2016 | ~0.04 | Risk-On / Low Vol |
| 2017 | ~0.08 | Risk-On / Low Vol |
| 2018 | ~0.15 | Stress / High Vol |
| 2019 | ~0.12 | Risk-On / Low Vol |
| 2020 | ~0.38 | Crisis then Recovery |
| 2021 | ~0.41 | Risk-On / Low Vol |
| 2022 | ~0.63 | Stress + Crisis |
| 2023 | ~0.32 | Risk-On recovery |
| 2024 | ~0.28 | Risk-On / Low Vol |
| 2025 | ~0.25 | Risk-On / Low Vol |
The structural break is visible: before 2020, correlation was negligible; from 2020 onward, it became materially positive and highly variable. The year-level averages obscure important within-year dynamics — particularly the spike-and-recovery pattern of crisis events — but the directional shift is unambiguous.
The Four-Regime Framework
This analysis classifies each month from January 2013 through December 2025 into one of four regimes defined by two observable variables: end-of-month VIX level and trailing 3-month S&P 500 total return.
| Regime | VIX | S&P 500 (Trailing 3M) | Example Periods |
|---|---|---|---|
| Risk-On / Low Vol | < 20 | Positive | 2017, 2019, 2021 H1, 2023–24 |
| Risk-Off / Low Vol | < 20 | Negative | Mild corrections, shallow pullbacks |
| Stress / High Vol | 20–35 | Either direction | 2018 Q4, early 2020 selloff |
| Crisis | > 35 | Either direction | March 2020, late 2022 extremes |
The framework is deliberately simple. Both variables are observable with a one-month signal lag, requiring no forecasting. The VIX threshold of 20 is the widely used boundary between complacent and elevated implied volatility regimes; 35 is a threshold historically associated with genuine market dislocations (financial crisis, pandemic, sovereign debt stress).
Average BTC-SPX rolling correlation and BTC return characteristics by regime, estimated across the 2013–2025 sample:
| Regime | Avg BTC-SPX 30-Day Corr | BTC Annualized Return | BTC Annualized Vol |
|---|---|---|---|
| Risk-On / Low Vol | ~0.25 | ~+85% | ~65% |
| Risk-Off / Low Vol | ~0.38 | ~+12% | ~72% |
| Stress / High Vol (VIX 20–35) | ~0.52 | ~-35% | ~90% |
| Crisis (VIX > 35) | ~0.68 | ~-48% | ~110% |
The monotonic increase in correlation as stress rises is the central finding. Bitcoin is most uncorrelated — and thus most valuable as a diversifier — in precisely the benign risk environment where equities are already performing well. In crisis regimes, correlation converges toward 0.68, meaning that Bitcoin and the S&P 500 move together closely enough that holding both provides limited risk reduction.
This pattern is consistent with Conlon and McGee's (2020) analysis of Bitcoin during the COVID bear market and with the broader post-institutionalization literature (Fang et al., 2022), which documented the structural correlation shift following the 2020–2021 institutional adoption wave.
Portfolio Impact by Regime
Across the full 2013–2025 sample, the portfolio comparison:
| Portfolio | CAGR | Annualized Vol | Sharpe Ratio | Max Drawdown |
|---|---|---|---|---|
| 100% S&P 500 | ~10.8% | ~17.5% | ~0.62 | ~-34% |
| 60/40 (S&P 500 + 10Y Bonds) | ~8.4% | ~11.5% | ~0.72 | ~-21% |
| 60/40 + 5% BTC | ~9.1% | ~12.1% | ~0.79 | ~-28% |
| 60/40 + 10% BTC | ~9.7% | ~13.8% | ~0.74 | ~-35% |
The 5% BTC allocation improves the full-sample Sharpe ratio of 60/40 from approximately 0.72 to 0.79. This reflects the sample composition: 2020–2021 was a dominant bull period for Bitcoin that contributed substantially to the full-sample average. The 10% allocation overshoots — higher CAGR but lower Sharpe and substantially deeper drawdown, owing to crisis-period amplification.
The same portfolios, evaluated specifically within each regime:
| Regime | 60/40 Sharpe | 60/40 + 5% BTC Sharpe | 60/40 + 10% BTC Sharpe | 60/40 Max DD | 60/40 + 10% BTC Max DD |
|---|---|---|---|---|---|
| Risk-On / Low Vol | ~0.94 | ~1.14 | ~1.28 | ~-8% | ~-9% |
| Risk-Off / Low Vol | ~0.38 | ~0.43 | ~0.44 | ~-12% | ~-15% |
| Stress / High Vol | ~0.31 | ~0.22 | ~0.14 | ~-18% | ~-26% |
| Crisis (VIX > 35) | ~0.18 | ~-0.05 | ~-0.19 | ~-21% | ~-38% |
The regime breakdown exposes what full-sample averages obscure. In Risk-On/Low-Vol environments, BTC allocations improved Sharpe ratios substantially — the diversification and upside capture benefits dominated. In Crisis regimes, however, a 10% BTC allocation produced a portfolio with roughly twice the maximum drawdown of plain 60/40 (-38% versus -21%) and a negative Sharpe ratio, meaning the risk-adjusted return was negative on average during those periods.
The specific crisis events in the data illustrate this concretely:
| Event | BTC Return | S&P 500 Return | BTC-SPX Correlation |
|---|---|---|---|
| COVID crash (Feb–Mar 2020) | ~-53% | ~-34% | ~+0.72 (spike) |
| COVID recovery (Apr–Dec 2020) | ~+303% | ~+65% | ~+0.15 (decoupled) |
| 2022 rate-hike bear market | ~-65% | ~-18% | ~+0.71 |
| FTX collapse (Nov 2022) | ~-22% | ~-5% | ~+0.68 |
| 2023 recovery | ~+155% | ~+26% | ~+0.32 |
The pattern recurs: correlation spikes during the stress event, then decays during the recovery. Portfolio managers who reduced crypto exposure during correlation spikes would have captured much of the 2020 and 2023 recovery upside while limiting the 2020 crash and 2022 drawdown amplification.
The Correlation-Monitoring Overlay
This observation suggests a systematic rule. Rather than a static Bitcoin allocation, a dynamic approach uses the observable correlation signal to determine exposure:
Entry/exit rule: When the trailing 90-day BTC-SPX rolling correlation exceeds 0.55, reduce BTC allocation to zero. Resume the position when correlation falls below 0.40. All signals are applied with a one-month lag to avoid look-ahead bias.
The 0.55 threshold sits between the Stress and Crisis regime averages (0.52 and 0.68 respectively), providing an early warning before correlation reaches peak crisis levels. The 0.40 re-entry threshold creates a hysteresis band that prevents excessive trading around the boundary.
During the 2013–2025 backtest period, the correlation signal triggered in approximately 18% of months — primarily concentrated in 2020 (March–May), 2022 (February–November), and the FTX collapse period. In Risk-On/Low-Vol regimes, the signal was almost never triggered, allowing full participation in Bitcoin's bull-market upside.
| Strategy | CAGR | Sharpe Ratio | Max Drawdown | Signal Triggered (% of Months) |
|---|---|---|---|---|
| 60/40 (no BTC) | ~8.4% | ~0.72 | ~-21% | N/A |
| 60/40 + 5% BTC (static) | ~9.1% | ~0.79 | ~-28% | N/A |
| 60/40 + 5% BTC (overlay) | ~9.3% | ~0.84 | ~-22% | ~18% |
The overlay improves Sharpe from 0.79 to 0.84 relative to the static 5% BTC allocation, while reducing maximum drawdown from approximately -28% to -22% — close to the plain 60/40 level. CAGR is slightly higher because the overlay exits BTC during its worst periods and re-enters during recoveries with restored full exposure.
The mechanism is intuitive: crisis regimes produce both high correlation (reducing diversification benefit) and large BTC losses simultaneously. The correlation signal serves as a coincident indicator of those regimes. By reducing exposure when the signal is elevated, the overlay systematically avoids the worst intersection of high correlation and deep BTC drawdowns.
This is consistent with Bouri et al.'s (2017) finding that Bitcoin's hedging and safe-haven properties are time-varying and instrument-specific, and with the broader literature suggesting that crypto's diversification benefits deteriorate precisely during equity market stress.
Limitations
Several limitations apply to this analysis.
The backtest period begins in mid-2013, when Bitcoin had sufficient trading volume and exchange infrastructure to constitute a replicable investment, but not spot ETF access. Pre-2024, obtaining Bitcoin exposure required either direct ownership (with custody risk), futures-based vehicles (with roll costs), or publicly traded holding companies (with premium/discount volatility). The cost of replication was substantially higher than the backtest assumes.
Bitcoin is used throughout as the proxy for "crypto." Ethereum has shown different, and historically higher, correlation with equities during stress periods. Altcoins vary widely. The backtest results do not generalize to crypto allocations beyond Bitcoin.
The correlation-monitoring overlay is defined and tested on the same 2013–2025 sample. The threshold values (0.55 entry, 0.40 exit), the lookback period (90 days), and the signal lag (one month) were all chosen with knowledge of the full historical record. This constitutes in-sample data-mining, and genuine out-of-sample performance will likely differ from the backtest. The overlay results should be interpreted as illustrative of the regime-conditional pattern, not as a validated forecasting system.
Macro regime overlap matters. The 2022 bear market combined high inflation (CPI > 8%) with rising rates and equity drawdown simultaneously — an unusual combination that may have amplified BTC-equity correlation beyond what either factor alone would produce. Extrapolating from a single dominant crisis event risks overfitting to that episode.
Tax treatment varies by jurisdiction and holding structure. In many markets, each quarterly rebalancing of the BTC allocation — and each overlay-triggered exit and re-entry — constitutes a taxable event. After-tax returns from an overlay strategy will differ substantially from the gross figures presented.
Practical Takeaways
The historical evidence from 2013 through 2025 supports several observations, presented in analytical rather than prescriptive terms.
Bitcoin's diversification benefit is regime-conditional. Across the full sample, BTC-SPX 30-day rolling correlation averages approximately 0.25 in Risk-On/Low-Vol regimes and approximately 0.68 in Crisis regimes. The unconditional average masks this variation and overstates the typical diversification benefit during adverse markets.
A static 5% BTC allocation has historically improved full-sample Sharpe ratios for 60/40 portfolios. The improvement (from ~0.72 to ~0.79) reflects the bull-market-dominated sample composition. Within crisis regimes specifically, the same allocation historically increased rather than reduced portfolio risk.
Correlation monitoring has historically provided a useful coincident signal. The 90-day BTC-SPX correlation crossing 0.55 has historically overlapped with the onset of Stress and Crisis regimes. The one-month signal lag means the overlay responds to confirmed regime shifts rather than attempting to predict them.
The overlay improvement is not large in absolute terms but structurally meaningful. Moving Sharpe from 0.79 to 0.84 while recovering drawdown from -28% to -22% represents a meaningful improvement in risk-adjusted efficiency for a 5% portfolio allocation. The complexity cost is modest: monitoring one rolling correlation metric and making allocation changes roughly three to four times per decade based on the historical trigger frequency.
Starting-period allocation matters. Investors who established BTC positions during peak-correlation periods — late 2021 and early 2022, when BTC-SPX correlation had already risen above 0.50 — experienced the worst risk-adjusted outcomes. The entry point into a Bitcoin allocation relative to prevailing correlation regimes has historically been as important as the allocation size itself.
Related
This analysis was synthesised from Quant Decoded Research by the QD Research Engine AI-Synthesised — Quant Decoded’s automated research platform — and reviewed by our editorial team for accuracy. Learn more about our methodology.
References
-
Bouri, E., Molnár, P., Azzi, G., Roubaud, D., & Hagfors, L. I. (2017). "On the hedge and safe haven properties of Bitcoin: Is it really more than a diversifier?" Finance Research Letters, 20, 192–198. https://doi.org/10.1016/j.frl.2016.09.025
-
Conlon, T., & McGee, R. (2020). "Safe haven or risky hazard? Bitcoin during the COVID-19 bear market." Finance Research Letters, 35, 101607. https://doi.org/10.1016/j.frl.2020.101607
-
Fang, L., Bouri, E., Gupta, R., & Roubaud, D. (2022). "Does global economic uncertainty matter for the volatility and hedging effectiveness of cryptocurrency?" International Review of Financial Analysis, 73, 101618. https://doi.org/10.1016/j.irfa.2020.101618
-
Makarov, I., & Schoar, A. (2020). "Trading and arbitrage in cryptocurrency markets." Journal of Financial Economics, 135(2), 293–319. https://doi.org/10.1016/j.jfineco.2019.07.001
-
Quant Decoded Research. (2026). "Factor Momentum Across Asset Classes: An Original Backtest." /en/factor-momentum-across-asset-classes-backtest