Rates and crypto: the regression nobody runs anymore

1. The thesis

For five years, the loudest macro thesis on Bitcoin was: BTC is digital gold; in a loose-money regime, BTC goes up; in a tight-money regime, BTC goes down. Plausible. Rhetorically tidy. Worth checking against the data.

2. What we measured

Daily BTC USD return regressed against:

• 10Y US Treasury yield change
• 2s10s spread change
• DXY change
• Fed-funds-implied rate-path change (Fed Funds futures)
• 5Y inflation breakeven change

Window: 2017-01-01 to 2026-03-01. Daily frequency. Standard OLS with HAC standard errors.

3. Five charts, summarised

Chart 1: 10Y yield vs. BTC return

Coefficient: −0.04. R²: 0.002. Statistically and economically zero. If yields are the BTC story, the story isn't in the daily data.

Chart 2: DXY vs. BTC return

Coefficient: −1.8. R²: 0.04. Statistically significant, economically modest. Strong dollar = soft BTC, but explains 4% of variance — the other 96% is something else.

Chart 3: BTC vs. NASDAQ daily return

Correlation: 0.36 over the full window. Tightening to 0.62 in 2022-2023 (the rate-hike phase), loosening back to 0.28 since 2024. BTC trades like a tech stock when rates move quickly. It trades like an idiosyncratic asset when rates are stable. The rate sensitivity isn't a constant — it's regime-dependent.

Chart 4: Halving cycle

The four-year halving cycle (post-halving year +1, year +2 peak, etc.) explains more variance than any rate variable in a horse race. Whether that's coincidence, market structure, or both is unsettled.

Chart 5: Liquidity proxy (Treasury General Account + bank reserves)

Coefficient: +0.42. R²: 0.07. Best single rate-related predictor we found. Net liquidity injection (broader than rates alone) correlates better with BTC than nominal rates do.

4. What this implies

The "rates story" for BTC is real but narrower than commonly claimed:

• Daily rate changes barely matter.
• Rate regime shifts matter — BTC behaves differently when Fed expectations are repricing rapidly.
• Net liquidity (Fed balance sheet, TGA, RRP) matters more than the headline rate.
• The crypto-native cycle (halving, derivatives positioning) matters more than rates over multi-month windows.

5. Caveats

OLS on daily data is the most honest first cut, but it ignores heteroskedasticity in fat-tail conditions. We re-ran the same regressions on weekly data, on log returns, and on rolling-windows; the punchlines didn't change. Code and data sources are in the appendix.

What this piece is NOT: a forecast. We're not predicting BTC direction from rates. We're saying the daily-rate-to-BTC story most macro Twitter trades is overrated, and the variables that do work (liquidity, regime) get less attention.

6. What we actually do with this

Our research desk doesn't trade BTC. We publish this kind of work because customers ask, and because it's the kind of analysis we'd want to read if we were on the other side of the screen. Use it as one input. Trade your own thesis.