Does Macro Actually Move Bitcoin?

01What data we used

• Span: July 2010 to September 2026, aligned to each Friday's close; mainly weekly, with monthly and daily data as supplements.
• Bitcoin: price, market cap, and MVRV (the ratio of market cap to holders' cost basis, a gauge of overall unrealized profit).
• Macro: US 10-year real rate, 2- and 10-year Treasury yields, the dollar index, the Fed's balance sheet and net liquidity, M2, and the assets of the five largest central banks.
• Cross-asset: Nasdaq, S&P 500, Russell 2000, the VIX, gold, oil, and inflation expectations.
• Credit: high-yield spreads, Moody's BAA corporate spreads, and the Chicago Fed financial conditions index.
• Crypto money: stablecoins (by coin and by chain) and the daily net flows of US spot Bitcoin ETFs.
• Validation: every conclusion went through a "coincidence test" (shuffle the data randomly several hundred times to see how large a relationship pure chance can produce) and was checked by finding the pattern on the first half of the data and testing it on the second. We ran the data locally from official sources and from public mirrors, and the results matched exactly.

02Three core findings

Finding 1: Bitcoin "plugged into" macro in two steps

Before 2018, macro had almost no effect on Bitcoin. Across 2010–2017, all macro and cross-asset factors combined explained less than "pure coincidence" would. The link formed in two steps:

• Step one, 2018: after CME Bitcoin futures launched, Bitcoin began moving with the dollar and gold.
• Step two, March 2020: around the COVID crash, Bitcoin's market cap settled above $100 billion and it began rising and falling with the Nasdaq.
• The 2024 launch of spot ETFs did not bring any new change.

Finding 2: Macro explains only a small part of Bitcoin's moves

Since 2022, the share of Bitcoin's weekly moves each factor can explain:

Stretched to monthly, explanatory power rises to 58%, with the Nasdaq still first (16.6%) and rates plus the dollar together at 5.5%, not statistically significant.

Why not look only at weekly data? Macro is a slow variable, and weekly tests understate it. So we also tested the slow scales: monthly; one quarter to six months; letting liquidity and M2 lead Bitcoin by 0–16 weeks; and comparing bear-market depth across five cycles. The results agree: at slower scales the Nasdaq (market optimism) and the cycle matter more, while rates, the dollar and M2 still don't rank near the top. Also, the monthly sample has only 56 observations and 12 variables; after penalizing for the number of variables (adjusted R²) it is 46%, and when we shuffle Bitcoin's data 500 times the same model can still produce about 38% by chance, so the part that truly beats coincidence is roughly 8% to 20%.

Is the Nasdaq "stealing" rates' explanatory power? No. Since 2022, weekly changes in real rates and weekly Nasdaq returns have moved in opposite directions (correlation −0.26), and every factor's variance inflation factor is below 2.3, far under the 5–10 at which collinearity usually becomes a worry. Real rates alone explain just 0.8% of Bitcoin's weekly moves; once the Nasdaq is added, rates add almost nothing. The decomposition method we use (Shapley) splits the overlap between two factors evenly, so it does not hand rates' whole share to the Nasdaq. "Hikes" and "stocks rising" happening together is a rare combination since 2022, which is why Question 1 uses a two-way scenario table instead of looking at each separately.

Finding 3: Cycle sets the direction, macro sets the tempo

Bitcoin's five bear markets fell −93%, −85%, −84% and −77%, and the current one −53% at its worst so far (taking June 30, 2026 as the provisional low). The depths follow the order of the cycles exactly and have nothing to do with the size of the rate shock. Even if this cycle makes a new low, the ordering won't change unless it falls below about $29,000 (−76.7% from the top, matching the last cycle).

• In 2022 real rates rose 2.8 percentage points and in 2018 only 0.6, yet 2022 fell less.
• In this 2025–26 bear market the Nasdaq actually rose 14%, and Bitcoin still fell 53% (so far).
• The three mature bull markets lasted almost exactly the same time (1,059–1,067 days), under wildly different macro conditions.

By horizon: within a week macro is more useful; at the one-quarter scale the cycle has a slight edge; at six months the two are about even.

03Ten retail questions

Read each question in the same order: first the common claim, then what the data says, and finally how to use it.

Q1Does a rate hike always mean a drop?

COMMON CLAIMRates go up, money flees into Treasuries, so Bitcoin must fall.

What the data says: only half right.

• Looking at weekly rate changes alone, there is almost no relationship with Bitcoin (correlation −0.09, not statistically significant).
• The current backdrop: the Fed began cutting in September 2024 and was still cutting into 2025; but in 2026 the economy turned strong, and on September 16 the Fed hiked again by 25 basis points to 3.75%–4.00%, its first hike since 2023. The 10-year real rate rose to 2.93%, the highest since 2008. So the "cutting cycle" is over, and a rate spike is the real environment now.
• Still, a real-rate spike of more than 0.4 percentage points within a quarter is an independent headwind: all else equal, Bitcoin gains roughly 15–20 percentage points less over the same period.
• What really decides the outcome is whether credit is loose. Since 2022, when real rates spiked:

Splitting by US stock optimism instead: when rates spiked but US stocks were upbeat, Bitcoin's median move over the same period was still +27%.

Q2Does Fed easing lift Bitcoin? Does "global liquidity leads Bitcoin by 2–3 months" hold up?

COMMON CLAIMM2 and global liquidity lead Bitcoin by about 10–12 weeks, so watching liquidity lets you position early.

What the data says: it doesn't hold.

• M2 and Bitcoin "correlate at 0.94" only because both curves trend up over the long run — a spurious correlation; measured on returns, the correlation is just 0.08.
• We tested 14 "who leads Bitcoin" pairings, finding the pattern on the first half of the data and testing it on the second; only 1 passed.
• The lead-lag relationship with global central bank assets flips sign from era to era.
• Outside research agrees: the widely cited 0.94 is a levels correlation, and its original author never claimed a fixed lead time; another survey found the lead drifting between 1 and 3 months.

You can separate "the price of money" from "the quantity of money": real rates are the price, and Fed net liquidity (total assets minus the Treasury's account and reverse repo) is the quantity. Since 2022, at the one-quarter scale net liquidity has had an independent positive effect, but it has clearly weakened since May 2024.

Q3Is Bitcoin digital gold, a safe haven?

COMMON CLAIMWhen crisis hits, money hides in Bitcoin.

What the data says: no.

• In the Nasdaq's worst weeks since 2022, the Nasdaq fell 5.0% a week on average and Bitcoin fell 6.0%, closing up in only 12% of those weeks; gold fell just 0.6% on average and closed up in 36% of them.
• Bitcoin falls with stocks but doesn't rise with them: when the Nasdaq falls 1%, Bitcoin falls 1.11% on average; when the Nasdaq rises 1%, Bitcoin rises only 0.39% on average.
• Outside academic research also finds that crypto assets and US stocks are "more tightly linked on the downside."

Q4Gold hits a new high — does the money spill over into Bitcoin?

COMMON CLAIMAfter gold runs, it's Bitcoin's turn: "digital gold" catches up.

What the data says: no stable evidence.

• We tested two common definitions of "gold at a high" (top 20% of six-month gains; near a one-year high). After either, Bitcoin's 13-week performance pointed the opposite way, and neither had an independent sample. Both definitions fail, which tells us there is only noise in there.
• From January 2025 to January 2026, gold rose 81% while Bitcoin fell 14%.
• The link between Bitcoin and gold existed only in 2018–2021 and has not been significant since 2022.

Q5Is a stronger dollar bearish?

COMMON CLAIMThe dollar index is a contrarian indicator for Bitcoin.

What the data says: right direction, weak force.

• Since 2022, the weekly correlation between the dollar index and Bitcoin is only −0.15, with explanatory power under 1%.
• This inverse relationship only began in 2018.

Q6Do ETF inflows mean institutions are bullish? Are big outflows a buy signal?

COMMON CLAIMETF creations are real money, "harder" than social-media sentiment, so sustained inflows are evidence of institutional bullishness; heavy outflows mean panic and a contrarian buying opportunity.

What the data says: the first half is right, the rest is wrong.

• ETFs are not just institutions. Public quarterly institutional holdings disclosures show institutions have always been a minority, with most holders being retail and small advisory accounts.
• Some of the money is hedge funds running a "basis trade": buy the ETF and short futures at the same time, earning the spread between the two, with no view on direction.
• ETFs chase price: a day's net inflow correlates at a high 0.60 with Bitcoin's move over the past 3 days, and almost not at all (0.04–0.07) with its move over the next 1–10 days. The same holds in 2024, 2025 and 2026.
• Heavy outflows are not a buy signal: in the 20% of weeks with the largest outflows, Bitcoin fell an average 3.5% over the following 4 weeks, versus a gain of 1.9% across all weeks. There was no rebound after heavy outflows.
• After weeks of heavy inflows, Bitcoin rose an average 6.6% over the next 4 weeks. The direction is "continuation," not "reversal," but the difference is not statistically significant, so treat it as a reference only.

Q7Stablecoin supply is growing — does that mean new money is coming?

COMMON CLAIMMoney enters crypto along the path "fiat → stablecoins → buy coins"; stablecoin growth happens at step one, so it shows money moving early.

What the data says: two problems.

First, price rises first and stablecoins grow after. Since the ETF launch:

Stablecoin growth is most strongly tied to past gains and unrelated to future gains. What actually happens is that people swap into stablecoins to enter after price has already risen.

Second, the stablecoin total contains a lot of money that isn't here to buy coins.

Not all USDT on Tron is remittances: retail users in Asia and Latin America often use it for OTC on-ramps, and exchanges also park large USDT balances on Tron. So part of it is money to buy coins and part is not, and on-chain data can't separate them. We show both bases side by side:

On either basis the conclusion is the same: no new money came in over the last quarter — flat including Tron, shrinking excluding Tron. The total is up 4% in a year, mostly from Tron and payments-institution stablecoins.

Also, tokenized US stocks are still small (about $2.5 billion in market cap, under 1% of stablecoins) and have almost no effect on the stablecoin data.

Q8ETFs are buying and stablecoins are shrinking — so who is selling?

What the data says: over the past 13 weeks, ETFs took in about $5.96 billion net; trading stablecoins excluding Tron fell about $4.56 billion, and including Tron they rose only about $0.5 billion.

That says this stretch was mostly existing holders changing venue, not a large wave of new money: crypto money on-chain did not grow, and buying came mainly through the ETF channel.

To be clear, what changed is the channel, not the crowd. As Question 6 showed, ETF buyers include plenty of retail and advisory accounts and also hedge funds running basis trades. So "existing holders changing venue" means money moving from on-chain and exchanges into ETFs held in US brokerage accounts. The buyers may be the same kind of people, but where price gets set has changed. Since 2024 the weekly correlation between ETF flows and Bitcoin's weekly move is 0.67, versus only about 0.3 for stablecoins: marginal pricing power now sits in the ETF channel.

How to tell new money from a venue change:

Q9Did rate hikes cause the 2022 crash?

COMMON CLAIMThe Fed's aggressive hikes knocked Bitcoin down.

What the data says: hikes were an amplifier, not the root cause. We broke the 2022 decline apart:

• The cycle was the root cause: the depth and duration of the 2022 decline fall within the range of the 2014 and 2018 bear markets.
• Hikes contributed about 31%–37% by dragging down US stocks.
• Crypto's own credit events contributed about 54%: the LUNA collapse, the Three Arrows Capital blow-up, and FTX's failure.
• In the half-year of the sharpest hikes (November 2021 to May 2022), Bitcoin actually fell more slowly than in past bear markets; the pace picked up with the credit events of May–June.

Q10How should you trade on FOMC day? Is a cut bullish?

COMMON CLAIMCuts are bullish and hikes bearish; on meeting day, trade with the direction.

What the data says: we looked at 7 key hike and cut events; Bitcoin's direction over the following 3–12 months didn't match the direction of policy. Seven is too few to prove FOMC meetings are unrelated to Bitcoin; what we can say is that the rule "cuts mean up, hikes mean down" has no support in history. After the same first hike, Bitcoin could rise 71% or fall 39% over the next 12 months; after cuts, it could rise 64% over 3 months (September 2024) or fall 26% (September 2025). Direction is set by where in the cycle you are.

BonusWhich futures premium should you use?

COMMON CLAIMMany sites use the premium on Deribit's quarterly contracts as a stand-in for the CME premium, calling it "the same measure."

What the data says: it's the same concept in different markets: Deribit is mostly crypto-native leveraged money, while CME is US institutions and basis-trade money, and their absolute levels often differ by several percentage points.

04What to watch now

Five daily readings (as of September 30, 2026)

05Claims you can stop believing

06Limits and disclaimer

• Correlation is not causation; most of the above are "happened together" relationships, not predictions.
• Cycle conclusions rest on only 3–5 cycles and are directional evidence. Event samples such as FOMC meetings (7) and gold highs are smaller still; the text uses them only to show that "no evidence supports a given claim," never to forecast.
• The low of the 2025–26 bear market (June 30, 2026) is the provisional low so far and still awaits confirmation by time; wherever the text uses it, it is marked "so far."
• Macro is a slow variable and weekly data alone understates it, so this piece also gives monthly, quarter-to-half-year and cross-cycle results; conclusions rest on what agrees across scales.
• How much of the stablecoin supply on Tron is money to buy coins can't be separated on-chain, so trading stablecoins are shown on two bases: including and excluding Tron.
• ETFs have only about two years and nine months of data, so some conclusions rest on limited samples; where results are not statistically significant, the text says so.
• This is research sharing, not investment advice.