Gold vs USD Cash —
Interest, Ownership,
and the 1980 Exception
Gold dominates compounded USD cash from every legally valid starting point. The interest adjustment doesn’t close the gap. The ownership ban doesn’t rescue the cash case. The 1980 exception is real, narrow, and regime-specific. The AI-generated counter-thesis that sparked this investigation was wrong on the data.
“The claim is simple — Gold has dominated the USD since the end of Bretton Woods in 1971 when analysing price only. The most important question we need to analyse is how true this claim is when adjusting for interest earned, and for the time period between 1944 and 1974 when the ownership of physical gold was illegal in the USA.”
This thesis arrived directly from the dataset originator. Tiaan Fourie supplied 15 sheets of monthly compound return data covering July 1944 to March 2026, then asked whether the two adjustments — interest reinvestment and the 1933–1974 gold ownership ban — materially alter the finding. That is the correct question. Both adjustments are analytically legitimate and frequently deployed by cash advocates to contest gold’s long-run dominance.
The claim itself is attributed to broader market consensus — Schoeman Rudman’s LinkedIn comment being the proximate trigger — rather than a personal assertion by Tiaan. He is a gold advocate testing the robustness of a position he holds, not defending one he has staked publicly. That distinction makes this investigation methodologically cleaner than Lab 007. The question is honest, the data is clean, and the proposed defeaters deserve a precise answer.
Note on the AI counter-thesis: in framing the submission, Tiaan ran the claim through Google’s AI Mode. The response generated stated that compound interest “frequently matches or outpaces” gold. That assertion is directly contradicted by Tiaan’s own dataset. Lab 008 tests the original thesis and — as a structural finding — addresses the AI-generated counter-claim that preceded it.
Five anchors.
Interest already in the data. No adjustment required.
The critical methodological point for this investigation: the USD cash series in Tiaan’s dataset is not a nominal dollar value. It represents total return — principal compounded monthly at the 3-month US Treasury Bill rate, with every coupon reinvested. When the cash series reads 1,118 from a base of 100, that number already reflects 51 years of compounded T-bill interest. The interest adjustment proposed as a defeater is not something to apply on top of the data. It is already in the data.
This matters because it means any claim that “interest eliminates gold’s lead” must be tested directly against the indexed figures — not against a nominal gold price with interest added as a separate calculation. The dataset does the work. The investigation reads the output.
Monthly compound return, fully reinvested
Total return — not nominal dollar value
Indexed to 100 at each anchor date
Interest is already embedded in every figure
This is what a dollar holder actually earned
Monthly close · Macrotrends/WGC basis
London Bullion Market
Indexed to 100 at each anchor date
No yield — total return = price return only
The zero-yield penalty is fully reflected
Aug 1971 — Bretton Woods collapse
Dec 1974 — Eve of US legal ownership
Dec 1977 — Mid-period neutral
Jan 1980 — Monthly avg $668, not $850 peak
Gold: $4,560/oz
Gold ATH: $5,286 (February 2026)
All five anchors run to this single endpoint
No selective termination
Dataset originator: Tiaan Fourie
Four anchors. Gold wins.
One anchor. Cash wins — barely.
First legal US ownership date
Post-Bretton Woods, full period
Only period cash wins
Out of five tested
| Anchor | Context | USD Cash (Mar 2026) | Gold (Mar 2026) | Cash p.a. | Gold p.a. | Winner |
|---|---|---|---|---|---|---|
| Jul 1944 | BW inception — gold fixed & illegal | 2,783 | 12,455 | 4.16% | 6.08% | Gold +192bps |
| Aug 1971 | BW collapse — Nixon shock | 1,377 | 10,510 | 4.92% | 8.90% | Gold 7.64× |
| Dec 1974 | First US legal ownership | 1,118 | 2,186 | 4.83% | 6.21% | Gold +138bps |
| Dec 1977 | Mid-period neutral anchor | 957 | 2,563 | 4.79% | 6.95% | Gold 2.68× |
| Jan 1980 | Monthly avg $668 — gold’s historic spike | 800 | 632 | 4.62% | 4.09% | Cash +53bps |
The chart makes the structure of the finding visible in a way the table cannot. Four rows show the amber bar consistently extending well beyond the green. One row reverses. That reversal — the 1980 exception — is the entire thesis of the cash case. It requires the most adversarial entry point available, and the gap it produces is 53 basis points per annum. Against that, the legal anchor (Dec 1974) produces a 138 basis point annual advantage for gold, compounded over 51 years into a 1.95× total return multiple.
The interest adjustment does not close the gap. It is already built into every number on this chart.
The adjustments are applied.
The gold case survives both.
After full interest reinvestment: from December 1974, gold returns 2,186 vs cash 1,118. From August 1971, gold returns 10,510 vs cash 1,377. From December 1977, gold returns 2,563 vs cash 957. The proposed defeater was already built into the denominator. Gold wins comprehensively from every legally valid starting point, interest included.
The only configuration where cash wins after full interest reinvestment is January 1980 — addressed in Finding F03.
Applying this constraint correctly does not rescue the cash case. The ownership ban removes the period most structurally unfavourable to gold — the Bretton Woods era, when gold was government-price-fixed at $35 while cash compounded freely. Eliminating that period and starting from January 1975 (gold at $176/oz, first legal ownership month) produces a comparison where gold returns 2,186 vs cash 1,118 over 51 years, after full interest reinvestment. The ownership constraint tightens the analytical framing. It does not change the winner.
The partial verdict reflects that the constraint is legitimate and should be applied — it just doesn’t do what its proponents argue. It is a correct methodological point with an incorrect conclusion attached.
What requires context: January 1980 was not a representative month. It was the peak of the most extreme gold price spike in recorded history — driven by the Hunt Brothers silver corner, Soviet invasion of Afghanistan, and Iranian hostage crisis simultaneously concentrating geopolitical fear premium into a single commodity. Selecting January 1980 as a starting point is equivalent to measuring equity returns from March 2000 or March 2021 — both technically valid, neither analytically representative.
The partial verdict reflects that the finding is genuine but context-dependent. The gap has closed dramatically — from a peak indexed spread of approximately 439 (cash) vs 41 (gold) circa the year 2000, to 800 vs 632 by March 2026. Gold’s ATH of $5,286 in February 2026 briefly brought the 1980 comparison to near-parity before the pullback to $4,560 widened it again. The directional trajectory favours continued compression.
The proposed defeater
was wrong on the data before the test ran.
“The Verdict: When long-term compound interest is reinvested into the USD, it frequently matches or outpaces physical gold’s price appreciation.”
When Tiaan submitted the thesis, he included an AI-generated analytical response from Google’s AI Mode that framed compound interest as a likely defeater of gold’s dominance. The response stated that interest reinvestment “frequently matches or outpaces” gold’s price appreciation. This is the claim the investigation was, in part, constructed to test.
The assertion is false. It is contradicted by the dataset Tiaan supplied alongside it. The cash series in that dataset already represents total return with full monthly T-bill reinvestment — the very adjustment the AI response treats as a corrective to apply. After applying it (by reading the dataset as constructed), gold wins 1.95× from the legal ownership anchor, 7.64× from 1971, and 2.68× from 1977. Cash wins from one starting point out of five. “Frequently matches or outpaces” is accurate for 20% of tested configurations.
This is not a marginal error. The AI response generated a confident verdict that directly contradicted the primary data source attached to the same conversation. The mechanism is straightforward: the model was reasoning from general financial commentary — where the compound interest argument against gold is widely circulated — rather than from the specific figures in front of it. The circulating claim is plausible enough to sound authoritative. It is not supported by the data at the legally valid anchor.
The Lab’s structural finding: when an AI-generated counter-thesis is deployed to frame an investigation before the data has been examined, and the counter-thesis is itself wrong, the investigation’s first job is to identify that. It does not change the verdict on the primary claim. But it materially changes what story the piece tells — from “does interest adjustment change the gold finding?” to “does a confident AI-generated counter-argument survive the dataset its own user provided?”
It does not.
The pattern here is analytically important beyond this single case. The compound interest argument against gold is widely enough circulated that it reads as settled knowledge. An AI trained on that body of commentary will reproduce it. But its validity is entirely anchor-dependent — it holds from January 1980, fails everywhere else — and the conditions under which it holds are precisely the conditions that require the most caveat (historic price spike, regime-specific macroeconomic environment, narrowing gap over time). A confident assertion that it “frequently” holds inverts the actual evidence ratio.
Jul 1944 – Mar 2026
Monthly compound returns
Gold spot vs 3M T-bill total return
Indexed to 100 at discrete anchors
Supplied by DM · 19 May 2026
London Bullion Market basis
WGC consolidated: Bloomberg,
Datastream, ICE BA, MCX India
Jan 1915 – Mar 2026
Lab 007 methodology applies
Secondary Market Rate
Monthly · Jan 1934 – Mar 2026
Federal Reserve Bank of St. Louis
Compounded monthly per period
Total return series — interest embedded
Partial.
Gold’s dominance survives both adjustments.
adjustment Fail
constraint Partial
exception Partial
thesis Structural
The claim — gold has dominated USD cash since 1971 on price only — is correct. The two proposed adjustments, interest reinvestment and the ownership ban, both fail to change that finding from the legally valid starting points.
The interest adjustment was already embedded in the dataset. After full monthly T-bill reinvestment, gold returns 2,186 vs cash 1,118 from the first legal ownership date (December 1974), 10,510 vs 1,377 from August 1971, and 2,563 vs 957 from December 1977. The interest adjustment does not close the gap. It was never the gap-closer its proponents claim, except from one starting point.
The ownership ban is a legitimate analytical constraint. Removing pre-1975 starting points from consideration eliminates the period during which gold was price-fixed and legally inaccessible — the period most structurally unfavourable to gold. Applying the constraint correctly, gold wins the legally valid comparison by 138 basis points per annum over 51 years.
The one genuine finding for the cash case is the January 1980 exception. Using the monthly average of $668 — the correct figure, not the intraday $850 peak addressed in Lab 007 — cash returns 4.62% p.a. vs gold’s 4.09% p.a., a 53 basis point annual edge over 46 years. This is real. It requires the most adversarial possible entry point — the single most extreme month in gold’s price history — and the gap has compressed from roughly 10:1 indexed at its peak circa 2000 to 1.27:1 by March 2026. The trajectory is clear.
The AI-generated counter-thesis that framed the investigation — that compound interest “frequently matches or outpaces” gold — is false for four of five periods tested, and false from every legally valid anchor. It is the kind of confident-sounding claim that circulates because it is plausible in summary form and wrong in the data. The Lab documents it as a structural finding for that reason.
