To measure how closely crypto tracks US equities, we matched BTC and ETH returns against US stock indices across more than 2,000 trading intervals from May 2018 to November 2026. Over the full period, both tokens moved with the S&P 500 at a moderate correlation of 0.29 for BTC and 0.31 for ETH.
Those averages cover a wide range. In the earliest observations, from 2018 and 2019, the correlation was slightly negative, while in 2026 it reached roughly 0.50 for both tokens. The Nasdaq-100 produces a similar picture.
Rolling BTC and ETH return correlations with the S&P 500
The relationship changes with the period. It strengthened around the pandemic and during the 2022 tightening cycle, weakened in 2023, and increased again. The event markers provide context rather than identifying causes.
The downside comparison is more concrete. The S&P 500’s worst 5% of returns comprised 107 observations with losses of at least 1.74%. In those intervals, BTC fell 81% of the time and ETH 79%, compared with about 49% for each token in the middle 90% of stock returns. Median returns were −2.6% for BTC, −3.8% for ETH and −2.5% for stocks.
Crypto returns and loss frequencies on large stock-market moves
The pattern survives removing March 2020 and recalculating the tail. Since 2022, using that period’s own worst-5% cutoff, BTC fell in 92% of observations and ETH in 95%. That recent tail contains only 61 observations, but the joint selling is more pronounced than in the full history.
Bitcoin Was Relatively Defensive Within Crypto
To compare the tokens directly, we measured the return of the ETH/BTC exchange rate. A negative return means ETH lost ground against BTC, whether both rose or both fell against dollars.
Since 2022, ETH underperformed BTC on 79% of stocks’ worst days. Its median ETH/BTC return was −1.52%, compared with −0.15% in ordinary equity conditions and +0.67% on stocks’ best days. The contrast between the worst and best groups is 2.19 percentage points.
ETH versus BTC on equity stress and rally days
The full sample gives a smaller version of the same pattern: ETH underperformed on 69% of the worst equity days, with a median ETH/BTC return of −1.01%, versus +0.42% on the best days. The difference persists after excluding March 2020. Resampling observations in blocks, to allow stress days to cluster, also leaves a negative worst-versus-best difference.
This is relative protection. On the recent sample’s worst equity days, BTC’s own median USD return was −3.6%, versus −5.6% for ETH. Holding BTC rather than ETH reduced the typical loss in those observations.
Crypto Price Variation Has a US-Hours Tilt
We then divided hourly crypto returns by trading window. To use complete hourly observations, we defined the US cash-session window as 10 a.m. to the actual stock-market close, normally 4 p.m. New York. This excludes the opening half-hour and accounts for shortened sessions.
From the January 2024 spot Bitcoin ETF launch milestone through the cutoff, these hours represented just 17.1% of observed time, but accounted for 33.0% of BTC’s realized variance and 30.0% of ETH’s. Realized variance here is the sum of squared hourly log returns: it measures price variation, regardless of direction.
Crypto’s concentration of price variation during US trading hours
The average absolute hourly move in that window was 2.09 times a weekend hour for BTC and 1.78 times for ETH. Removing each token’s biggest 1% of hourly moves leaves ratios near 2.05 and 1.75. The concentration therefore extends beyond a few exceptional jumps.
This is a finding about crypto’s own volatility timing. It does not mean one-third of its price variation was caused by stocks.
The Type of Episode Changes the Observed Relationship
Three illustrative windows show why a single correlation cannot describe every crisis:
During the FTX window, crypto’s losses coincided with an equity rally. But that week also contained the 10 November US inflation release. The divergence cannot establish that FTX had no effect on equities; other news was moving markets.
Likewise, the SVB comparison includes both the bank’s failure and the Sunday depositor-protection and funding response. BTC’s gain is not a clean measure of investors seeking an alternative to banks. The tariff window captures several days of announcement and escalation, ending before the 9 April policy modification.
These cases illustrate different outcomes: a crypto-specific failure, banking stress with a policy response, and a broad trade-policy selloff. They are selected examples with unequal durations and overlapping news. They help explain the limits of a general “crypto follows stocks” rule, without establishing which market transmitted a shock.
Crypto’s relationship with stocks depends on the token, the trading window and the shock. BTC often held up better than ETH during equity stress. US trading hours concentrated price variation, while individual crises sometimes sent crypto and equities in opposite directions. Correlation alone therefore offers an incomplete guide to diversification: what matters is how an asset behaves when the rest of a portfolio is under pressure. In this sample, crypto frequently shared that pressure, even when its longer-run correlation with stocks looked moderate.
