Research · Dataset

Gold and the money supply

"Gold rises because the money supply rises" is the most common macro explanation for gold, and the easiest to state wrongly. This page puts 801 months of gold and silver against US M1 and M2 and asks the question properly: not whether the levels move together — they do, but so does everything with a dollar sign — but whether money supply growth carries any information about gold returns, decade by decade, and whether trading it would have worked.

Data through 2026-09 · Gold source updated November 03, 2026 · M2 from fred.stlouisfed.org/series/M2SL

Source and method

Gold
World Bank Pink Sheet, monthly average gold price, column "Gold", US$ per troy ounce
Silver (control)
World Bank Pink Sheet, monthly average silver price, column "Silver"
M2
Federal Reserve Bank of St. Louis (FRED), series M2SL: Money Supply, M2, billions of US dollars, monthly, seasonally adjusted
M1
FRED series M1SL: Money Supply, M1, billions of US dollars, monthly
Inflation
FRED series CPIAUCSL: Consumer Price Index for All Urban Consumers, monthly
Coverage
801 months, 1960-01 to 2026-09; headline sample of 662 months from 1971-08
Download
gold-money-supply-monthly.csv (month, gold and silver prices, M2 and M1 levels, CPI, monthly returns, 12-month and 36-month changes for both gold and M2)
Source links
https://www.worldbank.org/en/research/commodity-markets · M2SL · M1SL · CPIAUCSL
Reproduction
scripts/gold_money_supply_analysis.py; independently recomputed from the published CSV by scripts/verify_gold_money_supply.py (144 checks)

Two readings of the same claim. The level question is whether the gold price and the money supply move together over the long run. The change question is whether this month's money supply growth predicts this month's or next year's gold return. Only the second is useful for anything, and the two answer very differently.

What a monthly return is. Gold and silver here are monthly average prices from the World Bank Pink Sheet, and money supply is a monthly stock measured at period end. A return is the change in a monthly average — not an entry and exit you could have executed. No transaction cost, financing or storage cost is modelled anywhere on this page.

Calendar arithmetic. Every change window is computed by named calendar month, never by row position. The CPI series has a hole at 2026/09 in the vintage used here, and positional indexing would silently turn some 12-month windows into 13-month ones.

Sample caveat. The headline sample starts in 1971-08. Gold was pegged to $35/oz in the US until August 1971, so earlier months are a policy artefact.

The headline: high in levels, absent in changes

Correlation, levels 0.907 log gold vs log M2, n = 662
Correlation, 12-month changes 0.044 p = 0.261, n = 650
R-squared on changes 0.0019 0.19% of the variation
Decades with a negative sign 4 of 6 the relationship is not stable

The correlation between the gold price and M2 in log levels is 0.907, with an R-squared of 0.8226. That number is the reason people believe the claim. It is also almost meaningless on its own: both series trend upward across a sample containing the 1970s inflation, the 2008 stimulus and the 2020 stimulus. Two trending series correlate strongly whether or not one causes the other.

Change the question to returns and the relationship is essentially absent. The correlation between 12-month gold returns and 12-month M2 growth is 0.044 (p = 0.261, n = 650), with an R-squared of 0.0019 — money supply growth accounts for about 0.19% of the variation in gold returns. On a month-to-month basis the correlation is 0.073 (p = 0.061), and over 36 months it is 0.042 (p = 0.291).

Why the level correlation is a trap

  • The control is silver, which has no scarcity story of its own and no central bank buying. Silver's level correlation with M2 is 0.764 — nearly as high as gold's. A high level correlation with money supply is therefore a property of the dollar and the metals, not of gold's scarcity narrative.
  • The level correlation is also mostly retrospective. M2 is measured at period end, so a large part of any 12-month M2 "growth" was already in the price before the month began.
  • None of this says money supply is irrelevant to gold. Real rates and the dollar are the channels that matter, and they are measured separately on the macro correlation page. This page tests one specific, popular claim.

Real money supply does not rescue the story

The sophisticated version of the claim says gold responds to real money supply — M2 deflated by inflation — because inflation erodes the dollar's purchasing power. This page tests it on exactly the same months as the nominal version, so the two numbers are directly comparable.

Measure 12-month correlation with gold Months
Nominal M2 growth 0.0456 649
Real M2 growth (CPI-deflated) -0.1482 649
M1 growth, for reference 0.0241 649

The third row is the correlation between M1 growth and silver, shown as a same-structure control.

Deflating by inflation makes the relationship worse, not better: the 12-month correlation with gold turns negative at -0.1482. If inflation is what drives gold, the inflation-adjusted version of the money supply should be the cleanest signal. It is the weakest.

Does money supply growth predict gold?

Split the sample by whether 12-month M2 growth was above or below the sample median of 6.08%, then look at what gold did over the following 12 months.

Gold after high M2 growth +9.07% 325 months
Gold after low M2 growth +7.61% 325 months
Difference +1.46 pp 95% bootstrap -1.90 to 4.63

The difference is +1.46 percentage points, with a 95% bootstrap interval of -1.90 to 4.63 percentage points. It crosses zero comfortably. Money supply growth does not sort subsequent gold returns into two distinguishable groups.

The lead-lag scan points the same way. Correlating gold's 12-month return against M2 growth at every lag from six months before to six months after, the largest absolute correlation is 0.0607 at a lag of -6 months. There is no lead-lag structure worth trading. For scale, gold's own 12-month autocorrelation is 0.958 and M2's is 0.986: both series are far more predictable from their own recent past than from each other.

Lag (months) Months Correlation
-6 644 0.0607
-5 645 0.0544
-4 646 0.0494
-3 647 0.0470
-2 648 0.0472
-1 649 0.0472
0 650 0.0442
+1 649 0.0330
+2 648 0.0202
+3 647 0.0094
+4 646 0.0006
+5 645 -0.0072
+6 644 -0.0143

The relationship changes sign by decade

If money supply really drove gold, the link would be positive in every period. It is not.

Decade Months 12-month correlation p Mean M2 growth Mean gold return
1970s 89 -0.5346 < 0.001 9.18% +24.54%
1980s 120 0.0380 0.680 7.62% +2.42%
1990s 120 -0.4381 < 0.001 3.92% -3.14%
2000s 120 -0.3619 < 0.001 6.24% +12.48%
2010s 120 -0.0345 0.708 5.67% +3.60%
2020s 81 0.0437 0.699 6.31% +17.21%

The correlation is negative in 4 of 6 decades, and in the 1970s, 1990s and 2000s it is negative and statistically significant — in the 1970s, the decade with the fastest money supply growth on record (9.18% a year), gold and money supply moved in opposite directions with a correlation of -0.535.

A relationship that changes sign is not a relationship you can lean on. It also explains the headline level correlation: the strong positive decades (1980s, 2020s) offset the strongly negative ones in the full-sample average.

Trading it: a walk-forward test

Statistics on a split can look tidy and still not be tradeable. The honest test is a rule with no look-ahead: each month, hold gold for the following month if 12-month M2 growth is above the median of everything observed so far, and hold cash otherwise. Walk-forward test starting after 60 months of history.

Money supply rule Just hold gold
Months 589 589
Total return -12.9% +2,878.6%
Annualised return -0.28% +7.16%
Maximum drawdown -51.8% -62.1%
Annualised volatility 8.1% 16.2%
Sharpe ratio (excess over cash) -0.487 0.258
Time invested 27.2% 100.0%

The rule lost money: -0.28% a year against +7.16% for holding gold the whole time, with a smaller drawdown only because it was out of the market 72.8% of the time. This is what a variable with no predictive content looks like when you trade it — a slightly different random walk.

Cash is assumed to earn 0% and gold is assumed to be holdable with no storage cost, no spread and no financing. Real-world costs would make the rule worse, not better.

2021 to 2026: the gap that broke the story

The most recent stretch is where the money supply story fails most visibly, and it is also the stretch most gold bulls will remember.

Window Gold M2 Silver Gold minus M2 Percentile since 1971
2021-12 → 2026-09 +141.3% +9.0% +187.1% +132.3 pp —
2021-12 → 2024-12 +47.9% -0.1% +36.9% +48.1 pp —
2023-12 → 2026-09 +113.2% +12.8% +170.3% +100.4 pp —
2020-12 → 2026-09 +132.5% +22.5% +158.4% +110.0 pp —
2019-12 → 2021-12 +21.0% +40.0% +31.6% -18.9 pp —
2008-12 → 2011-12 +101.0% +18.0% +194.2% +82.9 pp —

From 2021-12 to 2026-09 gold rose +141.3% while M2 rose +9.0% — a gap of +132.3 percentage points, the 93.6th percentile of every similar 48-month window since 1971. Silver rose +187.1% over the same period.

Money supply did not expand by anywhere near the amount that would be needed to explain the move, and the money supply was not growing fast enough at any point in the window to be the driver. Whatever moved gold and silver together here, it was not the quantity of dollars. Central bank buying, fiscal expectations and reserve diversification are the usual candidates; this page does not test them, and makes no claim about them.

The percentile compares the 2021-12 to 2026-09 gold-minus-money-supply gap with the gap in every other 48-month window in the sample, so it says how unusual this particular stretch is rather than asserting a probability.

The Bretton Woods years

The US official price of gold was fixed at $35/oz until August 1971. From 1968 the London market ran a two-tier system with an official price and a much higher free market price, so the Pink Sheet monthly average rises to $43/oz in this window. Either way, monthly moves here are a policy artefact rather than price discovery, and every headline statistic on this page excludes them.

In the file, the 1960-01 to 1971-07 window runs from 35.00 to 43.00 US dollars an ounce, with 36 months above the $35 official price as the two-tier London market opened up. Every headline statistic on this page excludes all 139 of them.

How to use this — and how not to

  • This page tests one claim: that money supply growth carries information about gold returns. On the evidence here it does not, and the relationship changes sign across decades. It is not a claim that monetary policy is irrelevant to gold — real rates and the dollar are the channels that carry the macro information, and they are measured on the macro correlation page.
  • A high level correlation between two trending series is not a causal finding. If you are using the M2 level as an argument for gold, the honest version of that argument needs a mechanism, and this page shows that money supply growth is not the mechanism. The same trap applies to official central bank gold buying: on the published IMF reserve data, the level correlation looks decisive and the change-on-change correlation does not — see central bank gold reserves and the gold price.
  • The walk-forward test uses monthly average prices and assumes cash earns nothing. It is a test of whether the variable has predictive content, not an executable backtest — no spread, storage, financing or tax is modelled.
  • Nothing on this page is investment advice, and no statistic here establishes a forward edge. Past correlations do not survive a change in monetary regime, and the 2020s sample is too small to extrapolate.

For the mechanism behind the real-rate and dollar relationships, see Gold, Real Yields and the Dollar. For the monetary-policy event study, see Gold and the Fed Cycle. For the policy-balance-sheet version of the same question — whether the Fed balance sheet itself carries the money story — see Gold and the Fed balance sheet. For how gold behaved in raw price terms, see Gold Price History.

Frequently asked questions

Does gold track the money supply?

Only in levels, and that is the least useful way to look. The log correlation between the gold price and M2 is 0.907 over 662 months, but both series trend upward, so that number mostly says "both went up". Measured on changes, the relationship largely disappears: the correlation between 12-month gold returns and 12-month M2 growth is 0.044 with p = 0.261 over 650 months, and R-squared of 0.0019.

If money supply does not explain gold, why does the correlation look so high?

Because both series have a strong upward trend over a sample that contains the 1970s inflation and the 2008 and 2020 stimulus waves. Trending series correlate strongly with each other whether or not one causes the other. The control on this page is silver, which rises with money supply for the same reasons gold does: the level correlation between silver and M2 is 0.764, so a high level correlation is a property of the money supply and the dollar, not of gold.

Does real money supply explain gold better than nominal money supply?

No, it explains it worse. Deflating M2 by CPI to get real money supply gives a 12-month correlation with gold of -0.148 over 649 months — the relationship turns slightly negative. On exactly the same months the nominal M2 correlation is 0.046, so the two are directly comparable. The inflation-adjusted version of the story is not stronger than the plain version.

Has money supply growth ever predicted gold?

Splitting the sample by whether 12-month M2 growth was above or below the sample median of 6.08%, the following 12-month gold return averaged 9.07% in the 325 high-growth months against 7.61% in the 325 low-growth months. The difference is +1.46 pp, with a 95% bootstrap interval from -1.9 to 4.63 percentage points, which crosses zero.

Does money supply lead gold, or does gold lead money supply?

Neither, in any measurable way. Scanning correlations from twelve months before to twelve months after, the largest absolute value is 0.0607 at a lag of -6 months. That is at or below the noise floor, so there is no lead-lag structure to trade here. For comparison, gold's own 12-month autocorrelation is 0.958 — a series is far more predictable from its own recent past than from money supply.

What did the 2021-2026 gold surge actually look like next to money supply?

From December 2021 to September 2026 gold rose 141.3% while M2 rose 9% — a gap of 132.3 percentage points, the 93.6th percentile of all similar 48-month windows since 1971. Silver rose 187.1% over the same period. Whatever drove gold and silver together here, it was not an expansion in the money supply.

Could you have made money by trading money supply growth?

The walk-forward test says no. Holding gold for a month whenever 12-month M2 growth was above the median of everything observed so far, and holding cash otherwise, produced a -0.28% annualised return with a 51.8% maximum drawdown over 589 months, against 7.16% and 62.1% for simply holding gold. The rule was in the market only 27.2% of the time.

How much of the sample is distorted by the Bretton Woods gold peg?

139 of the 801 months in the file, from 1960-01 to 1971-07, fall in the era when the US official price of gold was fixed at $35/oz. Monthly moves there are a policy artefact, not price discovery, so every headline statistic on this page uses only the 662 months from 1971-08 onward.

Educational content only. Not investment advice. Trading forex and CFDs carries substantial risk of loss.