Foresight has published a long-form study arguing that Bitcoin has not detached from the Federal Reserve. The link is still there, but it now works through a different transmission path.
The article re-runs data from Jan. 2017 to Sept. 15, 2026, covering 2,438 trading days and 62 Federal Open Market Committee meetings. BTC pricing is based on the BTC/USD daily close from Coinbase on FRED, macro variables come from official FRED series, and stablecoin data comes from DefiLlama. The author also states that the sample ends on Sept. 15, 2026, so intraday market action on Sept. 16 is outside the scope of the analysis.
The Sept. 16 rate hike did not trigger an instant BTC collapse
The article begins with the Sept. 16, 2026 Fed meeting, where the federal funds rate was raised by 25 basis points to 3.75% to 4.00%. It was the first rate hike in three years since July 2023. The decision passed 12-0, and in the first post-meeting press conference under new Chair Warsh, the dot plot showed that most officials still expected another hike later in the year.
That immediately revived a familiar market script: tighter liquidity should hit risk assets, and BTC should fall. But the article notes that Bitcoin did not show a classic policy-shock crash right after the statement. In the author’s view, that is not the most important point. The real question is how Fed policy reaches BTC prices over days, weeks, and years rather than in the first few minutes after an announcement.
The old “easing means BTC rises” formula failed over the past year
The article highlights a period that runs against the standard liquidity narrative. From Oct. 6, 2025 to Sept. 15, 2026, a span of 237 trading days, the effective federal funds rate fell from 4.09% to 3.63%, a 46-basis-point cut. The Fed balance sheet expanded from $6.59 trillion to $6.74 trillion, up $150 billion. M2 year-over-year growth rose from 4.19% to 4.59%, and net liquidity growth moved from -6.74% to +2.02%. Over the same period, the Nasdaq 100 gained 16.8% and the S&P 500 rose 13.0%.
Bitcoin moved the other way. It fell from $124,000 to $75,000, a drop of 38.2%.
The article says this is too large to dismiss as short-term noise. In the fourth quarter of 2025 alone, BTC fell 26.09%, while the S&P 500 rose 2.35% and the Nasdaq gained 2.31%. During that same quarter, M2 moved from 4.48% to 3.96%, the Fed balance sheet expanded from $6.59 trillion to $6.64 trillion, and the federal funds rate fell from 4.09% to 3.64%, a 45-basis-point cut. Rate cuts and balance-sheet expansion were both present, yet BTC did not rally.
The article then compares year-over-year correlations across regimes. In P1, the zero-rate and QE period from March 2020 to March 2022, the correlation between BTC year-over-year returns and M2 year-over-year growth was 0.716. In P2, the aggressive hiking period, it was still positive at 0.520. In P3, the high-rate period from July 2023 to the present, it flipped to -0.766. In other words, M2 accelerated while BTC fell.
The author adds an important qualification. This does not mean liquidity no longer matters. An Engle-Granger cointegration test on ln(BTC) and ln(M2) finds no cointegration in any of the three phases, with a p-value of 0.729. The article argues that the strong relationship seen in levels may simply reflect two trending series moving together without a long-run equilibrium.
The problem is not detachment. It is the observation window.
At the daily return level, the relationship between rates and BTC looks extremely weak. The article reports that the full-sample correlation between monthly M2 changes and BTC daily returns is -0.036, and the highest reading in any sub-period is only +0.031. The year-over-year change in the Fed balance sheet has a full-sample correlation of -0.029 with BTC daily returns. Daily changes in the federal funds rate show a full-sample correlation of -0.001 with BTC daily returns.
That can look like a disconnect, but the author argues that daily returns mix together expected hikes, unexpected hawkish signals, unexpected dovish signals, and ordinary trading days. Once those signals offset one another, the simple correlation naturally shrinks.
To address that, the article isolates FOMC dates and runs an event study. On FOMC days, the average daily move in the 2-year Treasury yield is 6.34 basis points, versus 3.89 basis points on non-FOMC days, a statistically significant difference with p=0.0015. Instead of using high-frequency target/path surprises, the study uses the daily change in the 2-year Treasury yield on FOMC days as a proxy for monetary-policy-related shocks. The author is explicit that this is only a proxy. It may also contain inflation expectations, growth expectations, term premium, press-conference information, and other macro news released that day.
The article then applies Jordà’s 2005 local projection method. For each horizon h, it estimates cumulative returns from the close before the FOMC meeting to h trading days later and regresses those returns on the 2-year yield shock proxy.
BTC reacts slowly to tighter Fed signals, then the move grows
Under the article’s daily-frequency setup, Bitcoin’s cumulative response on the FOMC trading day is close to zero. At h=0, beta is -0.11% and the t-statistic is -0.29, which is statistically indistinguishable from zero.
The author also flags a methodological limit. FOMC statements are released at 2:00 p.m. Eastern Time, while BTC trades 24/7. Because the study uses daily closing prices from FRED, it cannot cleanly identify the immediate high-frequency reaction after the announcement. So the result at h=0 means there is no clear response in the daily event window, not that the market showed no reaction at the instant of release.
The article references New York Fed staff report SR1052 by Benigno and Rosa, which described a “Bitcoin-Macro Disconnect” using intraday event studies and found no significant BTC response to macro news in their sample. The author says the h=0 result here points in the same direction, but the two should not be treated as identical because the frequency is different.
Once the horizon is extended, the picture changes. At h=1, BTC is still nearly flat at -0.03%. At h=5, the cumulative response reaches -1.43%. At h=10, it widens to -2.60%, with a t-statistic of -1.88, which the article describes as marginally significant. That move is 3.7 times the Nasdaq’s -0.70% and 5.8 times the S&P 500’s -0.45% over the same horizon. By h=20, the response narrows to -1.22%, and by h=30 it eases further to -0.54%.
The conclusion is straightforward: BTC is not unresponsive to tighter policy shocks. It responds later, and the move is larger.
After spot ETF approval, the delayed negative response becomes clearer
The article uses Jan. 11, 2024, the approval date for spot Bitcoin ETFs, as a structural break and examines the post-ETF subsample. The author notes that there are only 21 FOMC meetings after ETF approval, so the estimates should be read as directional evidence rather than precise point estimates.
Before ETFs, BTC’s coefficients across horizons are positive and insignificant, with h=30 even at +0.683, which the article interprets as broad immunity. After ETFs, the negative effect grows steadily with time: h=1 is -0.182, h=3 is -0.403, h=7 is -0.687, h=14 is -0.857 with t=-2.53, and h=30 is -1.063 with t=-2.07. The 30-day effect is 5.8 times the 1-day effect.
The article says this pattern looks less like a quick liquidation shock and more like valuation repricing that takes several weeks to work through the market.
Bonds price first, stocks later, BTC last but with a larger move
The study compares how different assets respond to the same tighter monetary-policy shock.
- On the first trading day, the 10-year Treasury yield rises by a statistically significant 1.71 basis points, with a t-statistic of 2.46. It is the only asset in the full sample to reach significance at h=1.
- Around the 20th trading day, the Nasdaq 100 and the S&P 500 reach statistical significance. The Nasdaq falls 0.82% with t=-2.16, and the S&P 500 falls 0.55% with t=-2.26.
- BTC reaches its peak response at h=10, with a cumulative move of -2.60%.
That places Bitcoin between bonds and equities in timing, but with a much larger magnitude.
The U.S. dollar index behaves differently. From h=0 to h=30, all eight horizons have absolute t-statistics below 1, with the largest at just 0.97. In the event-study framework used here, the article finds no significant evidence that the Fed transmits policy shocks to BTC through DXY in the short run.
The article also runs an exploratory Baron-Kenny mediation analysis. At h=0, the Nasdaq is the only significant candidate mediator, with a t-statistic of 2.35 and p=0.022. After controlling for the Nasdaq, the direct effect of the 2-year yield shock proxy on BTC shrinks from -0.107 to -0.071, a 34% reduction. The dollar is not significant as a mediator, with t=-1.63 and p=0.109, and the sign is opposite to what the simple narrative would suggest. The author says this is only consistent with an equity-market channel and does not prove a causal transmission chain on its own.
The dollar still often moves opposite BTC, but the dollar channel weakens after ETFs
The article separates two questions that are often treated as one. First, do the dollar and BTC often move in opposite directions? Second, does the Fed transmit policy shocks to BTC through the dollar?
For the first question, the answer is yes. The full-sample daily correlation between DXY and BTC is -0.197. Across the three phases, the correlations are -0.236, -0.281, and -0.097, all statistically significant. The article describes this as one of the more stable simple macro relationships across cycles.
For the second question, the event-study evidence is much weaker. As noted earlier, DXY does not respond significantly to tighter policy shocks at any horizon within 30 days after FOMC meetings.
The mediation analysis adds another layer. Before ETF approval, the Fed’s influence on BTC appears to run mainly through the dollar. A tighter policy shock lifts DXY, with a=+4.29 and t=4.87. A stronger dollar then weighs on BTC, with b=-0.035 and t=-3.60. The indirect effect is -0.152. During that period, the equity channel is not active in the event window.
After ETF approval, that dollar leg weakens sharply. Tighter policy shocks still lift DXY, with a=+3.73 and t=2.46, and they still raise real rates, with a=+0.296 and t=2.70. But DXY’s marginal effect on BTC falls from -0.035 to -0.002, and the indirect effect shrinks to -0.007. At the same time, a direct Fed-to-BTC channel appears, with a total effect of -0.182 and p=0.054. The article interprets this as evidence of a path that no longer needs to pass through the dollar, the Nasdaq, or real rates.
ETF approval changed market structure, but BTC did not become “another Nasdaq”
The article says the common assumption is that ETFs pulled Bitcoin into Wall Street and turned it into a more standard technology-style asset. The data, in the author’s telling, is more complicated.
When BTC daily returns are regressed on S&P 500 daily returns and Jan. 11, 2024 is used as the break date in a Chow test, the F-statistic is 1.51 with p=0.221. That means the null of structural stability cannot be rejected. The beta estimates on both sides of the break are noisy and insignificant: ETF pre-period beta is -0.130 with t=-1.41, and ETF post-period beta is +0.178 with t=1.45. The article explicitly says it would be wrong to claim that the S&P beta fell from 0.789 to 0.417 with statistical significance.
The Nasdaq 100 result is different. In the interaction regression, beta falls from 0.789 to 0.417, with the interaction term at t=-2.70 and p=0.0071. The Chow test gives F=3.944 and p=0.0195, which is statistically significant. Even there, the author warns against a simple story. The Nasdaq sample starts at a different date, and annual betas are highly non-monotonic: -0.371 in 2019, 1.176 during the aggressive hiking phase in 2022, 0.481 in 2023, nearly zero at 0.005 in 2025, and back to 0.805 in 2026.
The article argues that three structural changes are more robust.
- Volatility fell sharply. BTC annualized volatility dropped from 75.1% before ETF approval to 48.5% after, a decline of about 35%.
- The one-day lead effect from U.S. equities strengthened. Cross-correlation analysis shows that the predictive power of the previous trading day’s S&P 500 return for same-day BTC returns rose from 0.255 before ETFs to 0.379 after, with a stable peak at lag +1. The article says this lines up with Mohamad (2025), which found at the 5-minute frequency that ETFs led BTC price discovery about 85% of the time.
- The mean 90-day rolling BTC-SP500 correlation rose from 0.002 before ETFs to 0.062 after, with a Welch t-statistic of -11.14 and p<0.001.
Taken together, the article says institutions have made information flow between BTC and equities faster, price discovery more synchronized, and pure speculative noise less dominant. But that still does not make BTC a high-beta tech stock. One direct figure in the article is that, after ETF approval, a single-factor Nasdaq 100 regression has an R² of only 0.035, meaning the Nasdaq explains less than 4% of BTC volatility.
At the same time, BTC’s negative response to tighter monetary-policy shocks begins to accumulate over a two- to four-week horizon after ETFs, with a 30-day beta of -1.063. The article argues that macro sensitivity and stock beta are not the same thing. One asks whether policy shocks eventually show up in price. The other asks how tightly BTC co-moves with the Nasdaq on a day-to-day basis.
In several periods, BTC leads traditional risk indicators rather than follows them
The article says one of the most surprising findings came from a five-lag Granger causality test on daily returns.
In P1, the zero-rate and QE period from March 2020 to March 2022, the statistically significant direction is mostly from BTC to macro variables. BTC Granger-leads the S&P 500 with p=0.0015, VIX with p=0.0001, DXY with p=0.018, and the 10-year real yield with p<0.0001. The Nasdaq is bidirectional in P1: BTC to Nasdaq has p=0.012, while Nasdaq to BTC has p=0.042.
In P3, from July 2023 to the present, the same pattern returns strongly. BTC leads the S&P 500, Nasdaq 100, and VIX, all with p<0.0001. In that phase, the reverse direction from macro variables to BTC is not significant.
Only in P2, the aggressive hiking period from March 2022 to July 2023, do macro variables clearly lead BTC. The 10-year Treasury yield to BTC direction has p=0.0089, and the 10-year real yield to BTC direction has p=0.049.
The article stresses that Granger causality is not economic causality. It measures predictive lead-lag structure. Because BTC trades 24/7 while stocks and bonds trade only during weekday sessions, events in Asia hours or on weekends can be priced into BTC first and then show up in traditional markets later. That mechanical lead is real, and common shocks cannot be ruled out by Granger tests alone.
Even with those caveats, the author says the result challenges a popular habit in market analysis. Outside the aggressive hiking phase in 2022, the statistical basis for using macro variables to forecast BTC is thin. In many periods, BTC price changes come before moves in VIX, the S&P 500, and the Nasdaq in a predictive sense.
Stablecoins show a stronger long-run relationship with BTC than M2 does
The article then shifts to crypto-native liquidity. Earlier, it showed that ln(BTC) and ln(M2) do not cointegrate in any of the three phases, with p=0.729. Running the same Engle-Granger test on ln(BTC) and ln(total stablecoin market cap) produces a very different result: the test statistic is -4.160 and the p-value is 0.0042, indicating a statistical long-run equilibrium relationship.
The author is careful here as well. Cointegration does not prove that stablecoins cause BTC to move. The relationship could reflect common growth in market size, expansion of the crypto ecosystem, or broader adoption trends.
Still, the year-over-year relationship has strengthened sharply. The article reports that the correlation between stablecoin growth and BTC growth was -0.007 in 2024, essentially zero. It rose to 0.312 in 2025, then to 0.743 in 2026, and the rolling correlation over the most recent year reached 0.891. Over the same broad period, the correlation between M2 year-over-year growth and BTC year-over-year returns in P3 was -0.766. In the author’s framing, traditional liquidity and crypto-native liquidity are now sending opposite signals.
The article does not present stablecoins as a short-term trading trigger. At the weekly frequency, the relationship is weak and unstable. The correlation between weekly stablecoin changes and weekly BTC returns is only -0.093. In weekly Granger tests, the direction from stablecoins to BTC has p=0.0569, which is not significant. Lead-lag scans show a correlation of 0.201 at a lead of negative five weeks, meaning BTC appears to lead stablecoins instead. The correlation between 30-day stablecoin growth and BTC daily returns is also just -0.033. The article cites Oefele (2025), which found that ETF flows are more likely a result of price moves than a cause, with strong reverse causality.
The conclusion is that stablecoins behave more like a slow-moving variable that helps shape the valuation center of gravity rather than day-to-day price swings.
When stablecoin growth is strong, BTC trades more independently from equities
The article says the more informative result comes from a state-dependent regression. After splitting the sample by the pace of stablecoin growth, it finds that when stablecoins are expanding quickly, BTC’s beta to the S&P 500 is -0.220 with t=-2.10, a statistically significant negative relationship. When stablecoin growth stalls, BTC only weakly follows equities, with beta at +0.082 and not significant.
That leads to the article’s central framework: Bitcoin now faces two liquidity systems at the same time. One is traditional financial liquidity, including Fed rates, M2, the dollar, and equity risk appetite. The other is crypto’s own liquidity system, including stablecoin expansion and contraction, on-chain capital flows, and internal crypto risk appetite.
Within that framework, the article revisits the puzzle from Oct. 2025 to Sept. 2026. Traditional liquidity was easing, yet BTC fell. At the same time, total stablecoin market capitalization did not contract with BTC. It kept expanding, from above $300 billion to above $310 billion, reaching a record high. The author says this at least shows that BTC’s decline did not coincide with a parallel disappearance of crypto-dollar liquidity, and that traditional liquidity and crypto-internal liquidity may have diverged in timing.
The article also warns against overreading that point. Stablecoins can sit in wallets, on exchanges, in DeFi, in Treasury RWA products, idle, in arbitrage books, or in internal institutional transfers. A larger stablecoin supply does not mean all of that capital is waiting to buy BTC. The article also cites BIS working paper WP1219, which found that after tightening, stablecoin market cap falls while money market fund AUM rises. That suggests stablecoins respond to monetary policy slowly and cumulatively, not necessarily inside a 30-day FOMC event window.
Bitcoin is not one thing in every regime
The article’s final argument is that BTC cannot be captured by a single label.
It is not digital gold, the author says. On the 5% of trading days with the sharpest jumps in VIX, BTC’s average return is negative in all three phases, and the probability of a decline is above two-thirds in each. In P2, the aggressive hiking period, the probability of falling alongside the shock reaches 88.9%, with an average drop of 5.67%.
It is not simply another Nasdaq trade either. After ETF approval, a single-factor Nasdaq 100 model explains only 3.5% of BTC volatility. In the first quarter of 2025, BTC and the Nasdaq were even significantly negatively related, with beta at -0.377 and t=-2.47.
The article instead describes four faces of Bitcoin, each showing up under different conditions:
- Under tighter FOMC policy shocks, BTC behaves like a macro risk asset. After ETFs, tighter signals weigh on price over the following weeks, with a 30-day cumulative effect around the 1% scale.
- During extreme fear and VIX spikes, BTC acts as a high-beta amplifier. In the worst 5% of S&P trading days during P2, BTC fell 2.69 times as much as the S&P itself. In quantile regression, the left-tail beta is 2.391, or 2.2 times the median beta.
- When on-chain stablecoins are expanding rapidly, BTC behaves more like a crypto-native asset and decouples from equities, with beta at -0.220.
- On ordinary days without a clear policy shock, BTC can at times act as an early signal for global risk appetite because its 24/7 market structure lets it price hidden risks before stocks and VIX do, especially in the QE period and in P3.
The article says the better question is not whether BTC is a macro asset in general, but what kind of asset it is under a given state.
Methods and limits
The study says it does not rely on simple correlations alone. It uses event studies, local projections, Chow structural break tests, Baron-Kenny mediation analysis, Granger causality, cointegration tests, state-dependent grouping, quantile regression, and cross-correlation functions.
The author also lists several limits. The policy shock is proxied by the daily change in the 2-year Treasury yield on FOMC days rather than high-frequency USMPD target/path factors, so it may contain inflation expectations, term premium, and press-conference information. FOMC statements are released at 2:00 p.m. Eastern Time while BTC trades 24/7, so daily data cannot isolate the immediate post-announcement reaction. There are only 21 FOMC meetings in the post-ETF subsample. Daily ETF flow data and derivatives leverage data were not included. Mediation analysis can only show contemporaneous statistical direction, not a standalone causal chain. Cointegration is not causation, and Granger lead does not mean BTC “saw the future.”
The article closes with a narrower claim than many market narratives would suggest. It does not say the study proves what Bitcoin has become. It says nine years of data show that the old single-thread macro formula for BTC no longer explains the facts now on the table.
