Crypto Equities Are Not Levered Bitcoin
The claim worth testing
The convenient shorthand for crypto equities is that they are levered Bitcoin: buy Coinbase instead of BTC and get the same exposure with more torque. If that were true, two things would follow. Bitcoin would explain most of their daily moves, and their Bitcoin betas would be comfortably above one.
Both are testable with free data. This regresses daily log returns for Coinbase (COIN), Robinhood (HOOD) and Circle (CRCL) against Bitcoin over the last 251 trading days, then repeats the exercise controlling for the broad equity market, because a stock that moves with Bitcoin on risk-on days is not necessarily moving because of Bitcoin.
Bitcoin alone explains less than you would expect
| Ticker | Business | BTC beta | R² | Correlation |
|---|---|---|---|---|
| COIN | Exchange | 1.16 | 55% | 0.74 |
| HOOD | Brokerage | 0.92 | 33% | 0.58 |
| CRCL | Stablecoin issuer | 1.23 | 32% | 0.57 |
Coinbase is the most Bitcoin-driven of the three and Bitcoin still leaves 45% of its daily variance unexplained. For Robinhood and Circle, roughly two thirds of the movement has nothing to do with Bitcoin at all. Whatever these companies are, they are not a repackaged Bitcoin position.
Separating crypto from general risk appetite
Bitcoin and equities both sell off when investors get nervous, so a raw Bitcoin beta absorbs some ordinary market risk. Over this window Bitcoin and the S&P 500 are correlated 0.47, which is enough to matter and low enough that the two can be separated. Regressing on both at once splits the exposure.
| Ticker | BTC beta | SPY beta | R² both | R² SPY only | BTC adds |
|---|---|---|---|---|---|
| COIN | 0.98 | 1.28 | 60% | 29% | +31pp |
| HOOD | 0.65 | 1.96 | 44% | 31% | +13pp |
| CRCL | 1.05 | 1.28 | 35% | 16% | +18pp |
Robinhood is the clearest case of a company being misfiled. Its beta to the S&P is three times its beta to Bitcoin, and adding Bitcoin to a market-only model improves explanatory power by just 13 points. Robinhood is a high-beta retail equity that happens to have a crypto business inside it. The diversification its management describes is visible in the return data, not just the revenue mix.
The disconnect: Circle
Circle earns its revenue on the interest generated by the reserves backing USDC. That income is a function of stablecoin supply and short-term rates. Bitcoin's price is not an input. A first-principles view would predict Circle trades on the rate outlook and on USDC supply, with modest crypto sensitivity.
It does not. Circle's Bitcoin beta is 1.05, slightly higher than Coinbase's, and splitting the year in half shows the market's framing shifting fast.
| Ticker | BTC beta, first half | BTC beta, second half | SPY beta, first half | SPY beta, second half |
|---|---|---|---|---|
| COIN | 0.85 | 1.18 | 1.55 | 1.11 |
| HOOD | 0.68 | 0.57 | 2.06 | 1.94 |
| CRCL | 0.73 | 1.49 | 2.05 | 0.83 |
That gap between how a business earns and how its stock trades is the most useful thing in this data. It cuts both ways. If Circle really is a rate-and-supply business, a crypto drawdown that drags the stock down is repricing something that did not deteriorate. If the market is right that stablecoin supply is itself reflexive to crypto activity, then the revenue model is more crypto-exposed than its line items suggest, and the correct conclusion is that the income statement understates the cyclicality.
This report does not resolve which. It establishes that the two readings disagree, and that the disagreement has been growing.
What survives
Three findings hold up. First, none of these companies is levered Bitcoin; the best model here explains 60% of Coinbase's daily variance and less for the others, leaving 40% to 65% driven by company-specific factors. Second, Coinbase's Bitcoin beta is approximately one, so the leverage intuition is wrong in direction as well as degree. Third, the three businesses have genuinely different exposure profiles, which means treating "crypto equities" as a single category loses most of the information.
The practical implication is that these are not substitutes for spot Bitcoin, and they are not substitutes for each other.
Limits of this analysis
Data honesty notes
- Equity prices are daily closes from Yahoo Finance's chart endpoint, which is undocumented and unsupported. It is the only free source that returned a full year of daily history during this work: Finnhub's candle endpoint is paid-tier only, and Stooq's CSV download now sits behind a browser challenge. A production system should not depend on it.
- Bitcoin prices are daily closes from Binance.US, cross-checked against Coinbase and OKX spot at the time of writing (within 0.1%).
- Returns are daily log returns on the equity trading calendar. Bitcoin trades continuously; weekend moves are absorbed into the Monday return rather than dropped.
- Regressions are ordinary least squares computed directly from the price series, univariate against BTC and bivariate against BTC and SPY. No statistical package was used and no figure here is quoted from another source.
- Robustness. Bitcoin rose 22.7% in the five days to 2026-08-21, during the sample. Excluding those five days moves every beta by 0.03 or less, so the results are not an artifact of that move.
- Not tested: statistical significance of the beta differences, rolling betas at higher frequency, and whether the split-half drift is a trend or noise. Those require more history than one year provides.