Does Global Liquidity lead Bitcoin?

Green, Jack

BlockPhi Insights No. 1

Study29 October 2026Version 1

Does Global Liquidity lead Bitcoin?

Jack Green, founder

BlockPhi

Abstract

On the BlockPhi Global Liquidity Model, Global Liquidity’s six-week direction matched Bitcoin’s 11 weeks later in 224 of 345 weeks from January 2020 to October 2026 (65%, against about 52% by chance), and the lead profile peaks near 10 weeks. On BlockPhi’s earlier composite measure, 18 February 2022 to 20 March 2026, the association peaked at 11 weeks (coefficient 0.0404, t 3.13); on the Global Liquidity Model that single-lag coefficient is not significant, though Global Liquidity still precedes Bitcoin in a Granger test. This is a model estimate as of 28 October 2026, and past results do not indicate future results.

Key findings

Direction. Global Liquidity’s six-week direction, advanced 11 weeks, matched Bitcoin’s in 224 of 345 liquidity weeks from 3 January 2020 to 7 August 2026, with Bitcoin to 23 October 2026; a circular-shift test that keeps the overlapping windows gives p 0.003, and with currency translation taken out the match falls to 182 of 339 weeks (54%), close to chance (model estimates as of 28 October 2026).
Section 4 Figure 1 Table 1
Lead. The six-week lead profile peaks at 10 weeks (r 0.42 from 3 January 2020), with 11 weeks close behind, but a 13-week block bootstrap places the peak anywhere from 0 to 26 weeks: a range, not a transmission time (model estimates as of 28 October 2026).
Section 5 Figure 5
Effect. Net of Bitcoin’s own two weeks of momentum, the 11-week coefficient is 0.0022 (t 0.47) over 18 February 2022 to 20 March 2026; Global Liquidity’s weekly changes still precede Bitcoin’s returns in a Granger test from 3 January 2020 (p 0.002 to 0.005), while from 2022 the test runs both ways (model estimates as of 28 October 2026).
Section 6 Section 7 Table 2 Table 3
Earlier measure. On BlockPhi’s earlier composite measure, 18 February 2022 to 20 March 2026, the 11-week coefficient was 0.0404 (t 3.13, or 2.83 and 2.68 on the other two standard-error estimators) and the scan peaked at 11 weeks (0.197 against a band of 0.134); the published estimates reproduce to the printed decimal (model estimates as of 28 October 2026).
Section 8 Figure 7 Table 4

Model estimates as of 28 October 2026; past results do not indicate future results.

Data to: 23 October 2026
Measures: BlockPhi Global Liquidity Model; BlockPhi’s earlier composite measure
Exhibits: 8 figures, 4 tables; 6 figures open in the terminal, 4 with a free account
JEL: C32, E44, E51, G15
Keywords: Global Liquidity, Bitcoin, lead-lag, Granger causality, collateral

Two lines from May 2022 to October 2026 in the BlockPhi terminal: Global Liquidity’s six-week rate of change shifted 11 weeks forward, and Bitcoin’s six-week rate of change, each on its own axis. — Figure 1 Global Liquidity’s direction matched Bitcoin’s 11 weeks later in about two weeks of three

1 What does the Global Liquidity Model measure?

It measures how much collateral the world’s financial system has to lend against, and how far that collateral stretches. In BlockPhi’s own definition, as the Global Liquidity Model’s page states it:

The Global Liquidity Model is a composite that measures the balance-sheet capacity behind global credit: the balance sheets of five central banks plus 17 government debt markets, revalued every Friday in US dollars and multiplied by a collateral multiplier keyed to the US ten-year Treasury yield.

The terminal’s Global Liquidity page draws that construction as three nested layers in its Sources of Global Liquidity chart (Figure 2): the balance sheets of the Federal Reserve, the European Central Bank, the Bank of Japan, the Bank of England and the People’s Bank of China; the collateral base, which adds the government debt of 17 markets; and Global Liquidity, the base scaled by a collateral multiplier that models how many times the same collateral is pledged again once lenders have taken their haircuts. The idea follows the collateral re-use literature, where the Financial Stability Board (2017) and the Federal Reserve (Infante, Press and Strauss 2018) describe collateral multipliers measured from securities-financing data.

On 7 August 2026, the last week shown to a free account on 28 October 2026, the model put Global Liquidity at $194.89 trillion on a collateral base of $104.09 trillion, of which $25.89 trillion was central bank balance sheets and $78.20 trillion government debt. Model estimates; past results do not indicate future results.

Three nested areas from 2019 to August 2026 in the BlockPhi terminal: central bank balance sheets at the bottom, the collateral base around them, and Global Liquidity on top, in trillions of US dollars. — Figure 2 Central bank balance sheets sit inside the collateral base, and the base inside Global Liquidity

The methodology sets out how the multiplier was estimated, coverage lists the 22 components, and the explainer names the five sources of any week’s change: central banks, government borrowing, bond prices, the dollar and the multiplier.

2 Why not use Global M2?

Because Global M2 counts money that already exists as currency and deposits, while the Global Liquidity Model measures the collateral that new credit is raised against. Much of the funding in financial markets is secured lending, such as repo, which M2 does not count; Global M2 is not a part of Global Liquidity, and Global Liquidity against M2 sets the two side by side.

The word itself is contested. The Committee on the Global Financial System (2011) describes global liquidity as the ease of financing in global financial markets, and separates what central banks provide from what private lenders create. Global Liquidity has no single standard definition. The Global Liquidity Model is not directly comparable to global M2, to the BIS global liquidity indicators, to Fed net liquidity, or to proprietary indices, whose methods vary by vendor. A published lead of liquidity over Bitcoin is therefore a property of the series it was measured on, and every estimate here names its series and its dates.

3 Why would Bitcoin respond to Global Liquidity at all?

Because Bitcoin has no cash flows to discount, its price leans heavily on how much funding is available for risk, which is the quantity Global Liquidity tracks. Its supply follows a fixed schedule, so a change in demand shows up in price rather than in quantity.

Two margins help. The extensive margin, the slow widening of the holder base, sets the long-run trend; the intensive margin, the cyclical swing in how much holders, funds and lenders are willing to carry, is where funding conditions would act. A lead test is a test of the intensive margin, so every test below works on changes, not levels.

Funding liquidity and market liquidity reinforce each other (Brunnermeier and Pedersen 2009), and capital flows, credit and risky asset prices share a global financial cycle that moves with US monetary policy (Rey 2015). A common account runs from central bank balance sheets to bonds, then credit, then riskier assets, with Bitcoin among the last to receive the impulse; here that account is a hypothesis to test, not a finding.

The cycle in question is drawn in the Global Liquidity page’s growth chart (Figure 3), as 13-week growth annualised and as 52-week growth. In the week of 7 August 2026 the model’s 12-month growth read 6.51% and its 3-month annualised growth 1.13%. Model estimates; past results do not indicate future results.

Two growth lines from 2019 to August 2026 in the BlockPhi terminal: Global Liquidity’s 3-month annualised growth, the more volatile, and its 12-month growth, around a zero line. — Figure 3 Global Liquidity’s growth rises and falls in cycles

4 How often does Global Liquidity’s direction match Bitcoin’s?

In about 65% of weeks on the Global Liquidity Model, against about 52% by chance. Global Liquidity’s six-week change, advanced 11 weeks, matched the direction of Bitcoin’s six-week change in 224 of 345 weeks for liquidity weeks from 3 January 2020 to 7 August 2026, against Bitcoin prices to 23 October 2026. Over the 214 weeks of BlockPhi’s earlier study, 18 February 2022 to 20 March 2026, it matched in 140 (65%). Model estimates as of 28 October 2026; past results do not indicate future results.

In the terminal, this is Bitcoin Overview’s GL (+11W) & BTC 6W Changes chart (Figure 1, at the head of this article), a side card on that page: Global Liquidity’s six-week rate of change, shifted 11 weeks forward, against Bitcoin’s six-week rate of change (the card’s BTC 6W RoC, computed over 42 days of daily prices). The test counts the weeks in which the two sit on the same side of zero, using Bitcoin’s six-week log change on Friday closes, for liquidity weeks to 7 August 2026. Model estimates; past results do not indicate future results.

The six-week windows overlap, so neighbouring weeks are not independent and a plain count overstates the evidence. A circular-shift test keeps the overlap: it slides Bitcoin against Global Liquidity by every offset of at least 26 weeks, wrapping the end round to the start, and asks how often a shifted pairing matches as often as the true one. For the 345 weeks to 7 August 2026 it gives p 0.003. On the collateral base alone the match is 225 of 345 weeks (65%), and on BlockPhi’s earlier composite measure, over the earlier study’s weeks, 135 of 214 (63%), as Table 1 sets out. Model estimates as of 28 October 2026; past results do not indicate future results.¹

Much of the direction match runs through the dollar. With currency translation taken out of Global Liquidity, so that a stronger or weaker dollar no longer moves the total, the same test matches 182 of 339 weeks (54%) from 14 February 2020 to 7 August 2026, against about 53% by chance (circular-shift p 0.47). The dollar-neutral series still has a lead profile that peaks, at 7 weeks (r 0.26, circular-shift p 0.007), so something besides the dollar remains. Does the dollar lead Bitcoin? takes up that channel. Model estimates as of 28 October 2026; past results do not indicate future results.

Table 1 The six-week direction test, by series and window Weeks; shares and chance in per cent; p from a circular-shift test

Series	Liquidity weeks	Weeks	Same direction	Share (%)	Chance (%)	Circular-shift p
Global Liquidity, on the Global Liquidity Model	3 January 2020 to 7 August 2026	345	224	64.9	51.6	0.003
Global Liquidity, on the Global Liquidity Model	18 February 2022 to 20 March 2026	214	140	65.4	50.1	0.012
The collateral base, on the Global Liquidity Model	3 January 2020 to 7 August 2026	345	225	65.2	51.2	0.003
The collateral base, on the Global Liquidity Model	18 February 2022 to 20 March 2026	214	141	65.9	50.1	0.006
BlockPhi’s earlier composite measure	18 February 2022 to 20 March 2026	214	135	63.1	50.2	below 0.01
Global Liquidity with currency translation removed, on the Global Liquidity Model	14 February 2020 to 7 August 2026	339	182	53.7	53.2	0.47

Note: The sign of the liquidity series’ six-week change against the sign of Bitcoin’s six-week log change 11 weeks later, Bitcoin to 23 October 2026, re-estimated on 28 October 2026 (the earlier measure’s row on its own sample). Chance is the rate two unrelated series with the same shares of rising weeks would match. Weeks before 31 July 2026 are reconstructed history. Model estimates; past results do not indicate future results.

5 How far does Global Liquidity lead Bitcoin?

By about 10 weeks on the Global Liquidity Model, with a wide margin either side. The six-week lead profile, re-estimated on 28 October 2026 on data to 23 October 2026, peaks at a lead of 10 weeks (r 0.42 from 3 January 2020, 0.41 from 18 February 2022), with 11 weeks close behind (0.41 and 0.38). Model estimates; past results do not indicate future results.

In the terminal the lead is something to look at before it is something to test. Lead & Forecast’s GL vs BTC Price chart (Figure 4) has a lag slider that moves Global Liquidity forward against Bitcoin’s price: at 11 weeks, Global Liquidity on 7 August 2026, $194.89 trillion, sits against Bitcoin’s close of $81,076.46 in the week of 23 October 2026. Two upward-trending series look aligned over a wide range of shifts, the trap Granger and Newbold (1974) called spurious regression, so the slider shows the idea and the tests use rates of change. Model estimates; past results do not indicate future results.

Bitcoin’s weekly price and Global Liquidity shifted 11 weeks forward, from 2022 to October 2026, on two axes in the BlockPhi terminal’s Lead & Forecast page, with the lag slider set at 11 weeks. — Figure 4 Two rising series look aligned at many shifts, so the lag slider reads the lead rather than tests it

Measured on changes, the picture is Figure 5: the correlation of Global Liquidity’s six-week change with Bitcoin’s six-week change k weeks later, for every k from 0 to 26. A 13-week block bootstrap of 2,000 resamples places the peak anywhere from 0 to 26 weeks, 95 times in 100, and the split-sample check in the terminal’s own lag scan does not settle on one lag: the held-out years peak at 11 weeks for data from 2020 and at 17 for data from 2022. On data to 23 October 2026 the lead is a range, not a transmission time, and the model itself is not a price forecast. Model estimates as of 28 October 2026; past results do not indicate future results.

A curve of correlation against lead, 0 to 26 weeks, rising to a peak at 10 weeks and falling below zero by 22 weeks, for data from 2020 and from 2022. — Figure 5 The six-week correlation peaks at 10 and 11 weeks, but the peak’s interval spans the whole scan

Most weekly tests point to the same neighbourhood. On the Global Liquidity Model, over 18 February 2022 to 20 March 2026, the weekly cross-correlation peaks at 10 weeks (0.152, against a 95% band of 0.134), and a vector autoregression’s impulse response of Bitcoin to a liquidity shock peaks at week 10 in all three specifications, as it does from 3 January 2020 to 23 October 2026; at 11 weeks itself the weekly correlation is 0.042, inside the band. Extended from 18 February 2022 to 23 October 2026, the impulse response peaks at week 2 instead, on a window where the reverse Granger test is significant too (p 0.012 to 0.026, Table 3). On BlockPhi’s earlier composite measure the same scan peaked at 11 weeks (0.197, Figure 7). Model estimates as of 28 October 2026; past results do not indicate future results.²

6 How large is the effect, net of Bitcoin’s own momentum?

On the Global Liquidity Model, the single 11-week coefficient is not statistically significant. Regressing Bitcoin’s weekly log return on Global Liquidity’s weekly change 11 weeks earlier, with ARIMA(2,0,0) errors for Bitcoin’s own two weeks of momentum, gives 0.0022 (standard error 0.0046, t 0.47) over 18 February 2022 to 20 March 2026, re-estimated on 28 October 2026 (Table 2). Model estimates; past results do not indicate future results.

On BlockPhi’s earlier composite measure, over the same weeks of 18 February 2022 to 20 March 2026, the coefficient was 0.0404 (standard error 0.0129, t 3.13): a one trillion dollar weekly rise in that measure was associated with a weekly Bitcoin return about 4.1% higher 11 weeks later, a unit illustration rather than a scenario. Bitcoin’s own momentum terms, -0.0445 and 0.0485, were not significant. From 3 January 2020 to 23 October 2026 the Global Liquidity Model’s coefficient is 0.0069, with t of 1.87, 1.54 or 2.32 depending on the standard-error estimator, so whether t clears 2 turns on a choice the data cannot make. Model estimates; past results do not indicate future results.

One likely reason the weekly and six-week tests part company can be read off the Global Liquidity page’s History panel (Figure 6). The model revalues the collateral base in US dollars every Friday, reprices government debt for bond yields in six of its markets, and takes in new debt as each quarter of BIS figures arrives, so one week’s change mixes the dollar, bond prices and quarterly steps with the slower movement of balance sheets: between 10 January 2020 and 7 August 2026 the collateral base’s largest weekly rise was $4,759.4 billion (20 November 2020) and its largest fall $1,848.6 billion (30 September 2022). The earlier measure and the Global Liquidity Model track each other closely in level and over six weeks, far less week to week, and a weekly regression is dominated by such single weeks. Model estimates; past results do not indicate future results.

Weekly bars of the change in the collateral base from January 2020 to August 2026 in the BlockPhi terminal, in billions of US dollars, rising and falling around zero. — Figure 6 One week’s change in the collateral base mixes balance sheets, new debt and revaluation

Table 2 The weekly cross-correlation and the single 11-week coefficient, by window Leads in weeks; correlations; the coefficient, its standard error and RMSE in weekly log return

Window	Cross-correlation peak (weeks)	r at the peak	r at 11 weeks	95% band	Lag-11 coefficient	Standard error	t (Hessian)	t (outer product)	t (robust)	RMSE
18 February 2022 to 20 March 2026	10	0.152	0.042	0.134	0.0022	0.0046	0.47	0.43	0.52	0.072
18 February 2022 to 23 October 2026	9	0.135	0.07	0.125	0.004	0.0041	0.97	0.84	1.11	0.07
3 January 2020 to 23 October 2026	10	0.133	0.094	0.104	0.0069	0.0037	1.87	1.54	2.32	0.079

Note: On the Global Liquidity Model, Bitcoin’s weekly log return regressed on Global Liquidity’s weekly change 11 weeks earlier with ARIMA(2,0,0) errors, t under three standard-error estimators, re-estimated on 28 October 2026. Weeks before 31 July 2026 are reconstructed history. Model estimates; past results do not indicate future results.

7 Does Bitcoin lead Global Liquidity instead?

Not on BlockPhi’s earlier measure; on the Global Liquidity Model the evidence runs one way from 2020 and both ways from 2022. Global Liquidity’s weekly changes precede Bitcoin’s returns in a Granger test with p of 0.01 to 0.02 over 18 February 2022 to 20 March 2026, and 0.002 to 0.005 from 3 January 2020 to 23 October 2026, re-estimated on 28 October 2026. Model estimates; past results do not indicate future results.

The reverse test, that Bitcoin precedes Global Liquidity, is not clean on the data from 18 February 2022 (p 0.04 to 0.13 on the window to 20 March 2026, and 0.012 to 0.026 on the window to 23 October 2026), so the two series show feedback there and no one-way claim is made; from 3 January 2020 it does not reject (p 0.23 to 0.40). On BlockPhi’s earlier composite measure, 18 February 2022 to 20 March 2026, the forward test gave p of 0.064, 0.080 and 0.036 with 11, 12 and 13 lags, and the reverse test p of 0.97 to 0.99: no evidence of precedence in that direction.³ Model estimates as of 28 October 2026; past results do not indicate future results.

Table 3 Granger tests in both directions and the impulse-response peak, by window Lags and the peak in weeks; p values from Granger tests

Window	VAR lags	Global Liquidity to Bitcoin (p)	Bitcoin to Global Liquidity (p)	Impulse-response peak (week)
18 February 2022 to 20 March 2026	11	0.012	0.08	10
18 February 2022 to 20 March 2026	12	0.013	0.128	10
18 February 2022 to 20 March 2026	13	0.02	0.04	10
18 February 2022 to 23 October 2026	11	0.006	0.016	2
18 February 2022 to 23 October 2026	12	0.011	0.026	2
18 February 2022 to 23 October 2026	13	0.016	0.012	2
3 January 2020 to 23 October 2026	11	0.002	0.283	10
3 January 2020 to 23 October 2026	12	0.005	0.4	10
3 January 2020 to 23 October 2026	13	0.005	0.235	10

Note: On the Global Liquidity Model, by window and number of weekly lags in the vector autoregression, re-estimated on 28 October 2026. A Granger test shows precedence, not cause. Weeks before 31 July 2026 are reconstructed history. Model estimates; past results do not indicate future results.

8 What did BlockPhi’s earlier measure show?

An 11-week lead, on weekly data from 18 February 2022 to 20 March 2026, and the estimates behind it reproduce. The measure is read as last published on 17 July 2026. Model estimates; past results do not indicate future results.

BlockPhi’s earlier composite measure is the series the firm used before the Global Liquidity Model replaced it, and the 11-week figure BlockPhi has quoted was estimated on it, over 18 February 2022 to 20 March 2026. Its numbers stay here, labelled, so that a quoted figure can be traced to the series it came from. Table 4 reproduces the full set in Python: the coefficient, its standard error and t, the AIC and both peaks match the earlier study, run in R, to the printed decimal, and the reverse Granger p values come out a little lower, at 0.97 to 0.99. Model estimates; past results do not indicate future results.

Figure 7 is that measure’s cross-correlation scan over 18 February 2022 to 20 March 2026: a peak of 0.197 at 11 weeks against a 95% band of 0.134, leads of 9 to 12 weeks above the band, and a circular-shift p for the whole scan below 0.006. Three vector autoregressions, with 11, 12 and 13 lags, place the peak of Bitcoin’s impulse response to a liquidity shock at week 12 in each; the autoregression was not told where to look, which gives that result more weight than the scan. The regression’s root mean squared error was 0.070 in weekly log return, its residual lag-1 autocorrelation 0.004 and its AIC -494.25. Model estimates; past results do not indicate future results.

Bars of correlation for leads from minus 20 to 20 weeks, with a band either side of zero; the bars at 9 to 12 weeks rise above it, the tallest at 11 weeks. — Figure 7 On the earlier measure, leads of 9 to 12 weeks cleared the significance band

The two series are not interchangeable: on the Global Liquidity Model, re-estimated on 28 October 2026, the 11-week coefficient is not significant, while the direction test, the Granger test and a lead profile near 10 weeks remain. Model estimates; past results do not indicate future results.

Table 4 The estimates on BlockPhi’s earlier composite measure, reproduced

Statistic	Estimate on the earlier measure
Cross-correlation peak	lead of 11 weeks, r 0.197 (band 0.134); leads of 9 to 12 weeks above the band
Granger, earlier measure to Bitcoin (VAR 11, 12, 13)	p 0.064 / p 0.080 / p 0.036
Granger, Bitcoin to the earlier measure (VAR 11, 12, 13)	p 0.971 / p 0.984 / p 0.990
Impulse-response peak (VAR 11, 12, 13)	week 12 / week 12 / week 12
Lag-11 coefficient (ARIMA(2,0,0) errors, no intercept)	0.0404, standard error 0.0129
t of the lag-11 coefficient	3.13 (numerical Hessian); 2.83 (outer product); 2.68 (robust)
Unit illustration	a one trillion dollar weekly rise associated with a weekly Bitcoin return about 4.1% higher 11 weeks later
Bitcoin’s own terms, AR(1) and AR(2)	-0.0445 (t -0.63) and 0.0485 (t 0.67)
AIC	-494.25
Root mean squared error (weekly log return)	0.070
Residual lag-1 autocorrelation; Ljung-Box (10 lags)	0.004; p 0.47
Weekly pairs in the regression	203 (liquidity from 18 February 2022, Bitcoin from 6 May 2022 to 20 March 2026)

Note: BlockPhi’s earlier composite measure of Global Liquidity, as last published on 17 July 2026, weekly data from 18 February 2022 to 20 March 2026 (214 weeks), reproduced on 28 October 2026 in Python. These are not results on the Global Liquidity Model. Model estimates; past results do not indicate future results.

9 How can the comparison be rebuilt on the Chart Desk?

With weekly Coinbase BTC-USD candles open on the Chart Desk, Global Liquidity is added from the watchlist as a Compare overlay on its own scale.

The terminal’s BTC vs Liquidity preset, in the Chart Desk’s strip on phones and tablets, builds the same layout in one press, with Binance’s BTCUSDT pair as the main symbol in place of Coinbase’s. Either way the chart can then be extended, rescaled and saved.

Figure 8 is that layout on 28 October 2026: weekly Coinbase candles over five years, with Global Liquidity as a Compare overlay on its own scale, ending on 7 August 2026 for a free account. Across the same window the article’s own price series, Bitstamp’s Friday close, ran from $16,508 (25 November 2022) to $122,308 (3 October 2025); the Chart Desk draws Coinbase’s weekly candles, whose closes differ from Bitstamp’s Friday closes. Global Liquidity rose from $164.99 trillion on 29 October 2021 to $194.89 trillion on 7 August 2026. Model estimates; past results do not indicate future results.

Weekly Coinbase Bitcoin candles from late 2021 to October 2026 on the BlockPhi Chart Desk, with Global Liquidity drawn over them as a line on its own left scale, ending in August 2026. — Figure 8 Unshifted, Bitcoin and Global Liquidity both rose over five years, at very different rates

The Chart Desk shifts neither series, so it complements the lag slider on Lead & Forecast rather than replacing it. Its price-scale menu sets a series on a logarithmic or percentage scale, and its FRED series, the Federal Reserve’s broad dollar index among them, can be laid over the same chart to follow the dollar’s part in the direction result.

Notes

1
A supporting check. This check was not chosen blind: a September 2026 run of the model had found 228 of 343 weeks on the collateral base to 24 July 2026, so the criteria for this re-estimation, written before any statistic was computed, named it a supporting check, not a primary one. On the same 343 weeks this run finds 225; the reconstructed history is unchanged since 30 July 2026, and the difference is a different read of Bitcoin’s weekly close. Model estimates as of 28 October 2026; past results do not indicate future results. Back to text
2
A moving lead. The lead need not stay put. Since US spot exchange-traded funds began trading in January 2024, Bitcoin can be held in the same brokerage accounts as Treasuries, which may shorten the path from funding to price; these samples do not test that, and the transmission lag is re-estimated rather than fixed. Back to text
3
Precedence. A Granger test shows precedence, not cause (Granger 1969). Back to text

References

Brunnermeier, M. K. and Pedersen, L. H. (2009). “Market Liquidity and Funding Liquidity.” Review of Financial Studies 22(6). doi.org/10.1093/rfs/hhn098 (opens in a new tab)
Committee on the Global Financial System (2011). “Global liquidity: concept, measurement and policy implications (CGFS Papers No 45).” Bank for International Settlements. bis.org (opens in a new tab)
Financial Stability Board (2017). “Non-Cash Collateral Re-Use: Measure and Metrics.” Financial Stability Board. fsb.org (opens in a new tab)
Granger, C. W. J. (1969). “Investigating Causal Relations by Econometric Models and Cross-spectral Methods.” Econometrica 37(3). doi.org/10.2307/1912791 (opens in a new tab)
Granger, C. W. J. and Newbold, P. (1974). “Spurious regressions in econometrics.” Journal of Econometrics 2(2). doi.org/10.1016/0304-4076(74)90034-7 (opens in a new tab)
Infante, S., Press, C. and Strauss, J. (2018). “The Ins and Outs of Collateral Re-use (FEDS Notes).” Board of Governors of the Federal Reserve System. federalreserve.gov (opens in a new tab)
Künsch, H. R. (1989). “The Jackknife and the Bootstrap for General Stationary Observations.” Annals of Statistics 17(3). doi.org/10.1214/aos/1176347265 (opens in a new tab)
Ljung, G. M. and Box, G. E. P. (1978). “On a measure of lack of fit in time series models.” Biometrika 65(2). doi.org/10.1093/biomet/65.2.297 (opens in a new tab)
Rey, H. (2015). “Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy Independence (Working Paper 21162).” National Bureau of Economic Research. nber.org/papers/w21162 (opens in a new tab)
Seabold, S. and Perktold, J. (2010). “statsmodels: Econometric and statistical modeling with Python.” Proceedings of the 9th Python in Science Conference. doi.org/10.25080/Majora-92bf1922-011 (opens in a new tab)
Sims, C. A. (1980). “Macroeconomics and Reality.” Econometrica 48(1). doi.org/10.2307/1912017 (opens in a new tab)

Appendix A

How was it estimated?

From weekly data, Friday to Friday, with the tests and pass criteria fixed before any statistic was computed. The run was made in Python with statsmodels (Seabold and Perktold 2010) on 28 October 2026.

The data are the Global Liquidity Model as published on 23 October 2026, weekly from 24 May 2019; BlockPhi’s earlier composite measure as last published on 17 July 2026; and Bitcoin as Bitstamp’s daily close, the last of each Friday week, which matches the Bitcoin prices the terminal stores.

Recorded live since 31 July 2026. Weeks before 31 July 2026 are reconstructed with the current method, each input entering on a lag that follows its usual release schedule. They are not a record of what was known on each date.

A reconstruction can admit information nobody had at the time, which would flatter a lead test; of the 345 liquidity weeks in the direction test, only those of 31 July and 7 August 2026 were recorded live. Model estimates as of 28 October 2026; past results do not indicate future results.

Bitcoin enters as its weekly log return. Liquidity enters as its weekly change in US dollars for the cross-correlation, the vector autoregressions and the regression, and as its six-week percentage change for the direction test and the lead profile, paired there with Bitcoin’s six-week log change. A lead of k weeks pairs liquidity in week t with Bitcoin in week t + k.

The cross-correlation scan covers leads of 1 to 20 weeks against a band of 1.96 over the square root of the sample size. The vector autoregressions use 11, 12 and 13 lags with a constant, F-tests in both directions and an orthogonalised impulse response with liquidity ordered first (Sims 1980). The regression has ARIMA(2,0,0) errors and no intercept, as in the earlier study, with a Ljung-Box check of its residuals (Ljung and Box 1978); its standard errors are reported three ways, as the choice was not fixed in advance: the numerical Hessian (R’s arima convention, and the earlier study’s), the outer product of gradients and the robust sandwich. Over 18 February 2022 to 20 March 2026, t is 3.13, 2.83 and 2.68 on the earlier measure and 0.47, 0.43 and 0.52 on the Global Liquidity Model. Where windows overlap, the tests resample in a way that keeps the overlap: the circular shift by every offset of at least 26 weeks, and a 13-week moving-block bootstrap of 2,000 resamples (Künsch 1989), the method the terminal’s own lag scan uses. Model estimates as of 28 October 2026; past results do not indicate future results.

The 11-week lead would count as holding on the Global Liquidity Model only if, over 18 February 2022 to 20 March 2026, the scan peaked between 9 and 13 weeks above its band, the forward Granger test passed with a clean reverse test, and the 11-week coefficient had a t above 2. The first passed, at 10 weeks; the second failed on its reverse half; the third failed. On the earlier measure the published figures reproduced within tolerances set at the same time. Model estimates as of 28 October 2026; past results do not indicate future results.

Some things the method cannot separate. The samples span about two Bitcoin cycles; the tests are linear, while Bitcoin tends to fall faster than it rises; and neither a correlation nor a Granger test can tell transmission from a common cause, such as a move in US interest rates that reaches both series. Whether gold carries a lead of its own once Global Liquidity is held constant is tested in Does gold lead Bitcoin?. The re-estimation is scripted end to end, from the weekly data to every figure and table on this page, so each can be run again as the series moves on.

Appendix B

Data and charts

The charts in this article, live in the terminal: every figure here is a chart in the BlockPhi terminal or a BlockPhi calculation on a series the terminal draws, and Table B1 says where each one lives.

Figures marked as the earlier measure were estimated on BlockPhi’s earlier composite measure of Global Liquidity, as last published on 17 July 2026, using weekly data from 18 February 2022 to 20 March 2026 (214 weeks); every other figure was re-estimated on the BlockPhi Global Liquidity Model on 28 October 2026, with data to 23 October 2026. All are model estimates.

Table B1 Where each figure lives in the terminal

Figure	In the terminal	Access	Open
1	Bitcoin Overview > GL (+11W) & BTC 6W Changes	In the terminal for Premium members	Open Figure 1 in the terminal (opens in a new tab)
2	Global Liquidity > Sources of Global Liquidity	Free with an account	Open Figure 2 in the terminal (opens in a new tab)
3	Global Liquidity > Global Liquidity growth	Free with an account	Open Figure 3 in the terminal (opens in a new tab)
4	Lead & Forecast > GL vs BTC Price (Lag 11W)	Free with an account	Open Figure 4 in the terminal (opens in a new tab)
5	BlockPhi analysis of the BlockPhi Global Liquidity Model and Bitstamp Bitcoin closes, by the terminal’s lag-scan method	BlockPhi calculation	Open Figure 5 (Open Global Liquidity and Bitcoin in the terminal’s Lead & Forecast) (opens in a new tab)
6	Global Liquidity > Weekly change in the collateral base	In the terminal for Premium members	Open Figure 6 in the terminal (opens in a new tab)
7	BlockPhi analysis of its earlier composite measure of Global Liquidity, as last published on 17 July 2026, and Bitstamp Bitcoin closes	BlockPhi calculation
8	Chart Desk > BTC vs Liquidity	Free with an account	Open Figure 8 in the terminal (opens in a new tab)

Author

Jack Green is the founder of BlockPhi.

Versions

Version 129 October 2026 First published, with every statistic re-estimated on 28 October 2026 on data to 23 October 2026, and the earlier measure’s estimates reproduced and labelled.

Week	Global Liquidity (US$ tn)	Collateral base (US$ tn)	Central bank balance sheets (US$ tn)	Government debt markets (US$ tn)	Collateral multiplier
28 June 2019	121.19	71.94	20.27	51.67	1.685
27 September 2019	121.11	72.1	20.1	52.0	1.68
27 December 2019	122.81	72.78	20.57	52.21	1.687
27 March 2020	123.92	74.45	21.74	52.72	1.664
26 June 2020	133.64	80.21	25.5	54.71	1.666
25 September 2020	141.26	84.67	27.78	56.9	1.668
25 December 2020	154.8	92.22	29.72	62.5	1.679
26 March 2021	157.77	92.74	30.13	62.61	1.701
25 June 2021	162.41	95.49	31.17	64.33	1.701
24 September 2021	165.52	97.33	32.03	65.3	1.701
31 December 2021	167.65	98.15	32.51	65.63	1.708
25 March 2022	166.73	96.53	32.53	64.0	1.727
24 June 2022	161.02	91.62	30.83	60.79	1.757
30 September 2022	155.9	87.72	28.92	58.8	1.777
30 December 2022	162.13	91.03	29.36	61.68	1.781
31 March 2023	166.11	93.17	30.27	62.89	1.783
30 June 2023	164.04	91.45	29.0	62.44	1.794
29 September 2023	161.79	89.09	27.47	61.62	1.816
29 December 2023	168.48	93.03	27.93	65.1	1.811
29 March 2024	168.3	92.5	27.38	65.12	1.819
28 June 2024	167.43	91.72	26.08	65.64	1.825
27 September 2024	175.79	96.93	26.97	69.96	1.814
27 December 2024	170.65	92.87	25.63	67.24	1.838
28 March 2025	176.52	96.01	26.09	69.91	1.839
27 June 2025	183.95	99.65	26.35	73.3	1.846
26 September 2025	187.21	101.66	26.22	75.43	1.842
26 December 2025	187.13	101.29	26.0	75.29	1.847
27 March 2026	192.26	103.63	26.3	77.33	1.855
26 June 2026	194.72	104.38	26.16	78.22	1.865
7 August 2026	194.89	104.09	25.89	78.2	1.872

Week	3-month annualised (%)	12-month (%)
26 June 2020	35.24	10.27
25 September 2020	24.83	16.63
25 December 2020	44.22	26.05
26 March 2021	7.89	27.31
25 June 2021	12.31	21.53
24 September 2021	7.87	17.18
31 December 2021	6.28	8.5
25 March 2022	-2.49	5.68
24 June 2022	-13.01	-0.86
30 September 2022	-14.32	-5.58
30 December 2022	16.98	-3.29
31 March 2023	10.18	-0.12
30 June 2023	-4.88	1.24
29 September 2023	-5.39	3.78
29 December 2023	17.6	3.92
29 March 2024	-0.42	1.32
28 June 2024	-2.06	2.06
27 September 2024	21.54	8.66
27 December 2024	-11.2	1.29
28 March 2025	14.49	4.89
27 June 2025	17.92	9.87
26 September 2025	7.29	6.5
26 December 2025	-0.19	9.65
27 March 2026	11.43	8.91
26 June 2026	5.22	5.86
7 August 2026	1.13	6.51

Bitcoin week	Bitcoin close (US$)	Liquidity week (11 weeks earlier)	Global Liquidity (US$ tn)
25 March 2022	44320.6	7 January 2022	167.4
24 June 2022	21215.42	8 April 2022	166.86
30 September 2022	19425.0	15 July 2022	160.5
30 December 2022	16603.0	14 October 2022	155.93
31 March 2023	28476.0	13 January 2023	163.63
30 June 2023	30469.0	14 April 2023	166.09
29 September 2023	26911.0	14 July 2023	162.62
29 December 2023	42072.0	13 October 2023	160.99
29 March 2024	69903.0	12 January 2024	167.33
28 June 2024	60315.0	12 April 2024	167.47
27 September 2024	65781.0	12 July 2024	167.56
27 December 2024	94168.0	11 October 2024	174.08
28 March 2025	84379.0	10 January 2025	170.15
27 June 2025	107100.0	11 April 2025	178.39
26 September 2025	109696.0	11 July 2025	184.75
26 December 2025	87296.0	10 October 2025	187.66
27 March 2026	66359.0	9 January 2026	188.33
26 June 2026	60019.38	10 April 2026	192.87
25 September 2026	84089.54	10 July 2026	194.31
23 October 2026	81076.46	7 August 2026	194.89

Lead k (weeks)	r, from 3 January 2020	r, from 18 February 2022
0	0.161	0.111
1	0.136	0.082
2	0.142	0.039
3	0.147	0.005
4	0.169	0.02
5	0.203	0.068
6	0.289	0.159
7	0.352	0.228
8	0.389	0.315
9	0.412	0.395
10	0.425	0.414
11	0.41	0.378
12	0.362	0.312
13	0.309	0.271
14	0.251	0.199
15	0.212	0.115
16	0.174	0.07
17	0.133	0.034
18	0.106	0.006
19	0.099	-0.034
20	0.075	-0.083
21	0.034	-0.109
22	-0.026	-0.171
23	-0.054	-0.212
24	-0.094	-0.256
25	-0.147	-0.277
26	-0.175	-0.271

Year	Weeks	Net change (US$ bn)	Average weekly move, either way (US$ bn)	Largest rise (US$ bn)	Week of largest rise	Largest fall (US$ bn)	Week of largest fall
2020	51	19262.0	580.2	4759.4	20 November 2020	-1253.6	20 March 2020
2021	53	5924.8	397.6	2555.9	19 February 2021	-799.4	12 March 2021
2022	52	-7111.5	651.9	2570.2	18 November 2022	-1848.6	30 September 2022
2023	52	1997.3	440.3	1501.0	21 July 2023	-971.6	7 July 2023
2024	52	-164.3	510.6	1662.8	23 August 2024	-1431.9	11 October 2024
2025	52	8427.2	444.2	1894.9	21 February 2025	-608.6	18 July 2025
2026	32	2792.5	451.2	3249.3	20 February 2026	-803.2	13 March 2026

Does Global Liquidity lead Bitcoin?

Abstract

Key findings

1 What does the Global Liquidity Model measure?

2 Why not use Global M2?

3 Why would Bitcoin respond to Global Liquidity at all?

4 How often does Global Liquidity’s direction match Bitcoin’s?

5 How far does Global Liquidity lead Bitcoin?

6 How large is the effect, net of Bitcoin’s own momentum?

7 Does Bitcoin lead Global Liquidity instead?

8 What did BlockPhi’s earlier measure show?

9 How can the comparison be rebuilt on the Chart Desk?

Notes

References

Appendix A

How was it estimated?

Appendix B

Data and charts

Author

Versions

Also in BlockPhi Insights

Does the dollar lead Bitcoin?

Does gold lead Bitcoin?

What caused the 10 October 2025 crypto crash?

The model, live in the terminal.

Year	First liquidity week	Last liquidity week	Weeks	Same direction	Share (%)
2020	3 January 2020	25 December 2020	52	36	69.2
2021	1 January 2021	31 December 2021	53	33	62.3
2022	7 January 2022	30 December 2022	52	34	65.4
2023	6 January 2023	29 December 2023	52	33	63.5
2024	5 January 2024	27 December 2024	52	36	69.2
2025	3 January 2025	26 December 2025	52	34	65.4
2026	2 January 2026	7 August 2026	32	18	56.2
All weeks	3 January 2020	7 August 2026	345	224	64.9

Lead k (weeks)	Correlation	95% band	Position
-20	0.085	0.134	inside band
-19	0.087	0.134	inside band
-18	0.07	0.134	inside band
-17	0.056	0.134	inside band
-16	0.063	0.134	inside band
-15	0.073	0.134	inside band
-14	0.008	0.134	inside band
-13	-0.008	0.134	inside band
-12	0.01	0.134	inside band
-11	-0.015	0.134	inside band
-10	-0.053	0.134	inside band
-9	-0.034	0.134	inside band
-8	-0.008	0.134	inside band
-7	-0.034	0.134	inside band
-6	-0.026	0.134	inside band
-5	0.011	0.134	inside band
-4	0.015	0.134	inside band
-3	0.001	0.134	inside band
-2	0.021	0.134	inside band
-1	0.082	0.134	inside band
0	0.068	0.134	inside band
1	0.06	0.134	inside band
2	0.088	0.134	inside band
3	0.101	0.134	inside band
4	0.055	0.134	inside band
5	0.029	0.134	inside band
6	0.024	0.134	inside band
7	0.064	0.134	inside band
8	0.125	0.134	inside band
9	0.135	0.134	above band
10	0.161	0.134	above band
11	0.197	0.134	above band
12	0.17	0.134	above band
13	0.107	0.134	inside band
14	0.128	0.134	inside band
15	0.126	0.134	inside band
16	0.116	0.134	inside band
17	0.056	0.134	inside band
18	0.048	0.134	inside band
19	0.011	0.134	inside band
20	-0.028	0.134	inside band

Week	Bitcoin close, Bitstamp (US$)	Global Liquidity overlay (US$ tn)
29 October 2021	62301.59	164.99
31 December 2021	46214.37	167.65
25 March 2022	44320.6	166.73
24 June 2022	21215.42	161.02
30 September 2022	19425.0	155.9
25 November 2022	16508.0	161.85
30 December 2022	16603.0	162.13
31 March 2023	28476.0	166.11
30 June 2023	30469.0	164.04
29 September 2023	26911.0	161.79
29 December 2023	42072.0	168.48
29 March 2024	69903.0	168.3
28 June 2024	60315.0	167.43
27 September 2024	65781.0	175.79
27 December 2024	94168.0	170.65
28 March 2025	84379.0	176.52
27 June 2025	107100.0	183.95
26 September 2025	109696.0	187.21
3 October 2025	122308.0	186.59
26 December 2025	87296.0	187.13
27 March 2026	66359.0	192.26
26 June 2026	60019.38	194.72
7 August 2026	64877.77	194.89
25 September 2026	84089.54	not drawn (after the free view ends)
23 October 2026	81076.46	not drawn (after the free view ends)