Inside the Asset Bubble · Indicator test 1
Does the money supply actually outgrow the real economy? Testing M2SL against GDPC1, quarterly, 1959 – 2026.
The claim survives, but not on the timeline the thesis currently implies. Money supply did not steadily outrun real output from the 1970s onward. For the first fifty years of the series it didn't outrun output at all.
M2 as a share of GDP was flat to falling from 1959 through 2007, bottoming in 1997. The break is 2008, and it has never reverted. That is a narrower claim than "since the late twentieth century", and a far harder one to argue with.
The ratio
This is the chart the thesis needs, and it is not the chart the thesis expects. If liquidity had been quietly inflating past output since the 1970s, this line would slope up for sixty years. It doesn't. It oscillates in a narrow band, 0.46 to 0.61, through Volcker, through the 1987 crash, through the savings-and-loan crisis, through Glass-Steagall's repeal, through the dot-com bust.
Then 2008 happens, and the band breaks. The ratio steps up and stays up. COVID drives it to 0.888, the highest reading in the entire series. Four years of the sharpest tightening since Volcker have pulled it back to 0.708, which is still higher than any quarter between 1959 and 2009.
M2 money stock as a share of nominal GDP, 1959 Q1 – 2026 Q2
Both figures nominal, so the ratio is unaffected by inflation. This is the inverse of M2 velocity.
Indexed comparison
The ratio hides how much separation opened up. Indexing both series to 1980 = 100 and deflating M2 by the GDP deflator puts them on the same footing: purchasing power against production.
The two tracks run close together for twenty-eight years. Real M2 is at 232 in 2008 and real GDP at 225, within 3% of each other after nearly three decades. By 2021 real M2 is at 481 against real GDP's 300. Roughly 96% of the peak 2021 gap opened after 2008, or 93% of the gap that still remains in 2026, and about half of that again after 2020.
Real M2 vs. real GDP, indexed to 1980 Q1 = 100
M2 deflated by the implicit GDP deflator, so both series are in chained 2017 dollars.
Structural breaks
Growth rates by period settle the argument era by era. Positive means real money supply grew faster than real output; negative means output grew faster. Three of the eight eras run negative, including the entire 1990s, when M2 grew at less than half the rate of the economy.
The COVID bar is not a variation in degree. Real M2 grew 11.1 points a year faster than real output across 2020–21. Nothing else in sixty-seven years is within 7 points of it.
Annualized real growth gap: real M2 minus real GDP, by era
Percentage points per year. Eras follow the structural breaks named in the thesis.
Rate of change
The level chart shows where the ratio went. The year-over-year change shows when it moved, and this is the most damning chart in the set, not because the ratio rises, but because of how narrowly its extremes are distributed in time.
Across 1960–2019 the ratio's annual change has a standard deviation of 3.1%. Only ten quarters in sixty-five years move more than three standard deviations. All ten fall inside three episodes: 2009, 2020–21, and 2023. There is no fourth cluster. Nothing in the oil shocks, nothing in Volcker, nothing in 1987, nothing in the dot-com bust reaches the threshold.
2020 Q2 alone moves 29.3%, more than three times the ±3σ threshold, and far beyond anything else in the record. (The benchmark distribution is fat-tailed, so read that as historical extremity, not as a normal-curve probability.) In a series this long, that is not a business cycle; it is a policy event.
Year-over-year change in the M2/GDP ratio, 1960 Q1 – 2026 Q1
Percent change from four quarters prior. Dashed lines mark ±3 standard deviations of the 1960–2019 distribution.
The same chart carries the counter-evidence. 2023 is a three-sigma cluster in the negative direction, and it contains the only outright year-over-year contraction in M2 in the entire series, five consecutive quarters, bottoming at −4.2%. Seventeen of the eighteen quarters since 2021 Q4 have moved the ratio down, for a cumulative −17.6%.
Discipline
Section X of the thesis asks for strong claims to be separated from interpretive ones. Here is that split for this indicator pair.
M2V exactly.Real M2 is down 9.1% from its 2021 peak, and the 2022–26 era gap is negative at −4.6 pp/yr. Liquidity is currently being withdrawn relative to output, 17 of the last 18 quarters, −17.6% cumulatively. In 2023, M2 fell outright year-over-year for five straight quarters, bottoming at −4.2%: the only sustained nominal contraction anywhere in the sixty-five-year record.
If the thesis is that the system cannot tolerate monetary tightening, the last four years are the counter-example, and they have to be addressed head-on rather than left out. The honest version is narrower and stronger: the ratio was forced down hard and still sits 31% above its pre-2008 mean. The floor moved, even after the most aggressive withdrawal in the data.
Method
A units problem had to be fixed first. M2SL is nominal and GDPC1 is real. Dividing one by the other produces a line that rises roughly tenfold across the sample purely because prices rose. It measures inflation, not liquidity. Nominal GDP is therefore the denominator, and M2 is separately deflated for the indexed chart. Both routes give an identical ratio, which is the arithmetic check.
Two independent nominal-GDP sources agree. The ratio was first built on FRED GDP, then rebuilt on NGDPSAXDCUSQ (supplied separately). Across the 305 overlapping quarters the two differ by at most 0.0001%, and every headline figure on this page reproduces to seven decimal places on either one.
A second units trap sits in that series. NGDPSAXDCUSQ is reported in millions per quarter, not billions at an annual rate. Divided into M2 as-is it returns 0.0029 instead of 0.708, wrong by a factor of 245. It is converted here as × 4 ÷ 1000.
Frequency. Monthly M2 was averaged within each calendar quarter rather than point-sampled, so a single month's spike can't drive a quarter.
Coverage. The merged panel runs 1959 Q1 – 2026 Q2, 270 quarters. GDP data before 1959 exists but has no M2 counterpart and is excluded.
A supplied derived table was audited, not trusted. The quarterly workbook (m2-vs-nominal-gdp-quarterly-analysis.csv, 269 rows) was recomputed column by column from the raw series. All eight columns reconcile to floating-point precision, largest discrepancy anywhere in the table 5.5e-12, and its velocity column matches FRED M2V to within that series' published rounding. Its year-over-year columns are the basis of the rate-of-change chart above.
Sources. M2SL, GDPC1 and NGDPSAXDCUSQ from supplied CSVs; FRED GDP and M2V retrieved at analysis time as cross-checks. All are Federal Reserve Bank of St. Louis series, seasonally adjusted. The ratio chart runs to 2026 Q2 on the FRED pull; NGDPSAXDCUSQ ends one quarter earlier at 2026 Q1, where the ratio is also 0.708.
| Era | Real M2 | Real GDP | Gap |
|---|
| Quarter | M2 ÷ GDP |
|---|
Think of the whole country as one big shop. "Output" is how much stuff the shop actually makes and sells in a year. "Money supply" is how many dollars are floating around for people to spend. Normally those two grow together, more dollars chasing more stuff, and the prices stay sensible. That is exactly what happened for the first fifty years of this record, right through the 1970s and the 1980s and the dot-com crash. The dollars never got ahead of the stuff.
Then 2008 happened, and again in 2020, and the government created an enormous number of new dollars very quickly to stop the economy from seizing up. The amount of stuff being made did not jump anywhere near as much. So now there are far more dollars sloshing around for every real thing the country produces than at any time before 2010, and even after four years of pulling money back out, it has not returned to normal. The interesting part is when it happened: not slowly over fifty years the way people assume, but in two sudden bursts, both of them emergencies.