Indicators · finding · v1 · 2026-10-09
Borrowing against bitcoin does not move with old-coin spending
The question
A common story says holders borrow against their bitcoin instead of selling it. If that were a visible market force, weeks when bitcoin-secured credit grows would coincide with less spending of old coins, or with a rise in long-term-holder supply. The question, and the result that would contradict it, were written down before anything was computed.
The finding
From 2022-09-19 to 2026-10-05 (212 weeks), weeks of bitcoin-secured credit growth coincide with neither less nor more old-coin spending, nor with a change in long-term-holder supply. Credit growth does move with the price.
| Weekly change in… | vs long-term-holder supply | vs old-coin spending | vs coin-days destroyed | vs price |
|---|---|---|---|---|
| Secured debt, all venues | −0.03 / −0.04 | −0.07 / −0.07 | −0.04 / −0.07 | +0.25 / +0.29 |
| Secured debt, flagged venues out | −0.06 / −0.06 | −0.03 / −0.09 | −0.04 / −0.07 | +0.29 / +0.31 |
| Secured collateral, flagged out | +0.01 / +0.03 | 0.00 / +0.04 | −0.04 / 0.00 | +0.17 / +0.18 |
| Pool debt, flagged out | +0.03 / −0.01 | −0.11 / −0.12 | −0.09 / −0.08 | +0.32 / +0.36 |
| Minted-stablecoin debt, flagged out | +0.03 / −0.03 | −0.06 / 0.00 | −0.08 / −0.02 | +0.26 / +0.24 |
| Lending-desk debt, flagged out | −0.05 / +0.04 | +0.04 / −0.01 | +0.03 / −0.05 | +0.07 / −0.11 |
"Flagged out" leaves out the venue-weeks the daily integrity check flagged; the result is the same with them in.
Each cell is Pearson / Spearman correlation, n = 212. Chance alone gives |r| up to about 0.14 at 95%. No on-chain pair clears that bar; every price pair outside the lending desks does.
How it holds up
Taken net of the week's price change, every on-chain pair is unchanged. No single year clears its own bar. Of 187 rolling 26-week windows, 9 pass — about what chance gives at 5%. The result is the same with or without the venue-weeks the daily integrity check flags, and with Spearman in place of Pearson. Neither chain series moves with price, so price is not hiding a link.
Method
Credit growth is the change over seven days across the venues read on both dates, so a venue entering coverage never counts as growth. Lending desks enter by debt only: their bitcoin collateral is partly derived from dollar books at the day's price, so it moves mechanically against the price.
The chain side comes from our own node's pass over every block: long-term-holder supply (coins unmoved for 155 days or more), coins over one year old spent, and coin-days destroyed, each summed over the seven days ending Monday, then logged and differenced. Price is the on-chain basket fixing. Weeks are compared at the same time only: no lags were tested, so a delayed effect is neither found nor ruled out.
Limits
One credit cycle, so this is observational. It does not show the story is false for any one holder, only that it leaves no weekly, market-level trace in these series. The credit side is what is read: 83–100% of secured collateral in any week (median 100%). Nothing here is a forecast or a signal.
The weekly changes behind it — the change over seven days in secured debt and collateral, across the venues read on both dates — are kept, every version, as published series (matched weekly change). The correlations are computed from them and never stored, so a correlation is never presented as a signal.