Value in three monies

Every asset indexed to its base year (=100), then held to a constant monetary unit — US dollars, then ounces of gold, then ounces of silver, at the price prevailing in each year. A number that climbs in dollars often sits flat or falls once the money itself stops moving.

The three money-planes (3D)

Each asset draws a line across three parallel planes — USD, gold, silver. Log vertical scale (the range from a home at ~16 to NVIDIA at ~2,000 needs it). Drag to rotate; click a legend entry to toggle an asset; hover any point for the exact index.

Gold / silver ratio

Ounces of silver to buy one ounce of gold. A high ratio = silver historically cheap versus gold; the classic hard-money gauge. Annual averages.

Breakdown — endpoint index (base year = 100)

Read across a row: dollars rose the most, gold least, silver between. Only NVIDIA rises in every money — genuine value capture, not debasement. The broad market, housing and commercial real estate are all down two-thirds or more in gold since 2000.

Gold / silver ratio — annual

Wages in three monies

The same lens, turned on labor. A year's pay — by occupation, by region, gig versus traditional — held to US dollars, then to ounces of gold, then to ounces of silver at the price prevailing in each year. Wages roughly doubled in dollars since 2000. In gold they fell to a fifth. Figures are representative BLS annual means (occupations, regions) and the federal statutory minimum; approximate and directional, denomination is an overlay lens not proof.

A year of work, in three monies (3D)

Five benchmark earners — top, middle, floor, and the minimum wage — indexed to their year-2000 value (=100) across the USD / gold / silver planes. Every line rises steeply in dollars and collapses toward the floor in gold. Drag to rotate; toggle a series in the legend.

What a year's wage buys in gold — 2024 vs 2000

Each bar = a year's pay measured in ounces of gold, as a share of its own year-2000 purchasing power (dashed line = 100 = held even with gold). Ordered worst-first. Not one occupation kept pace with gold; the lowest-paid lost the most ground, and the federal minimum wage lost most of all.

By region — dollars up, gold down, everywhere

Census-region annual mean wage, indexed to 2000 (=100), in dollars (rose ~2×) beside gold-ounces (fell to ~a quarter). The regional spread barely matters once the money is held constant.

Gig vs traditional — a year's earnings in metal (2024)

Ounces of gold bought by a year of traditional full-time work (median) versus gig driving — gross, then net of vehicle and fuel costs. Gig work is a post-2010 category, so this is a 2024 snapshot; net figures are estimates and vary widely by market and hours.

Who earns — composition of work over time

Share of US employment by super-group, 2000→2024. The growing slices are the ones that also held the most gold value (management/professional, computer & math, healthcare); the shrinking slices — sales & office, blue-collar — are where the gold-denominated wage fell hardest. Composition and repricing move together: the workforce is tilting toward the few roles that outran debasement.

Wages — endpoint index & ounces (2000=100)

Read the two gold columns: a year's minimum-wage work bought 38 ounces of gold in 2000 and 6 in 2024; the average job, 122 then 27. The USD-index column climbs past 190 for almost every row — the same wage, told in two monies, tells opposite stories.

What can actually be proven

Everything above is an overlay. "Wages fell in gold" is nearly a tautology — it only re-denominates, and it privileges gold, so a critic answers "gold bubbled." This section keeps only what survives every such objection, and certifies exactly how un-handwavable that core is. Three tiers of certainty; the wage figures are representative and rounded, so only directions, never magnitudes, are claimed as proven.

The robustness certificate

Because the levels are approximate, a bare ratio is not enough. Model each of the four inputs (labor 2000, labor 2024, asset 2000, asset 2024) as known only to within a uniform relative error, and push all four at once in the direction most favorable to "no decline." The breakdown error e* is the answer to: how wrong would every number have to be, simultaneously, to make the decline disappear? A claim is CERTIFIED only if e* clears the assumed data tolerance of 15%.

Bars past the dashed 15% line are un-handwavable even granting my numbers could each be off by that much: against gold, silver and equities a year of average labor buys 22-51% of its 2000 exchange value, and overturning that needs 17-36% error in every input at once. Against housing and commercial real estate labor also fell — but by too little (e* of 3-7%) to certify at this data quality, so it is stated, not proven.

Same test on the federal minimum wage is stronger still (it fell furthest): certified against gold, silver and the S&P; the two hard-money columns need >33% uniform error to overturn.

Deeper — where did the decline happen? (real wages vs asset inflation)

The invariance theorem says the ratio fell; it does not say where. Any labor:asset ratio factors exactly into two observable pieces — labor priced in the consumer basket (the real wage) times the consumer basket priced in the asset (asset inflation in wage-hours):

This split is model-dependent — it trusts the CPI, which is contested (hedonic and substitution adjustments; alternative indices show higher inflation). If true inflation were understated, more of the fall would be lost real wages and less would be asset inflation. So the decomposition sits below the theorem in certainty: illuminating, but resting on a disputed deflator. The total labor:asset decline uses no CPI at all and stays certified.

Deeper — is it just a cherry-picked endpoint?

Recompute labor's relative price for every start year → 2024, not only 2000. Cells show the 2024 value as a share of that start year (★ = certified, e* > 15%).

Honest reading: the decline is strongest and certified from a 2000 or 2007 base (and, for equities, 2013). Measured from the recent decade it is milder and not certified — gold and silver were already elevated by 2013, so labor roughly held or even gained against them since. This is therefore a long-horizon (quarter-century) claim, and it is stated as one — it is not a claim that labor collapsed against hard money over the last ten years.

Deeper — the theorem is machine-checked

The two algebraic identities are not asserted in prose; the build recomputes each one two independent ways in exact rational arithmetic (Python fractions, zero floating-point error) and checks they are identical:

Deeper — authoritative series close the gap

The 15% tolerance above was a stand-in for rough figures. Swap in named primary series — BLS median usual weekly earnings, LBMA gold & silver, S&P annual close, S&P CoreLogic Case-Shiller (constant-quality) and Census median home — each with its own measurement tolerance (how precisely the annual number is known), and run the certificate as exact interval arithmetic per input. R_worst = R₀·(1+t_w)/(1−t_w)·(1+t_a)/(1−t_a); certified iff R_worst<1.

With authoritative-precision data the tolerances shrink from 15% to 1-3%, and the earlier hold-outs cross the line: homes now certify too — even the raw median, where a year of labor buys 80% of the house it did in 2000 and the decline survives the joint measurement error. The certified basket is no longer just hard money and equities; it is every independent store of value measured here.

Homes, honestly measured

"Median home price" hides several things, so reprice a home in wage-hours under each honest lens — constant-quality (repeat-sales), raw median, price-per-square-foot, and the monthly mortgage carry:

The rent signal the headline lags — ALNRI

Renters feel the same measurement gap. The official CPI rent line above (BLS CUUR0000SEHA) is a ~1-year-stale lagging print: new-lease market indices — the Apartment List National Rent Index (ALNRI) and the BLS New Tenant Rent Index (NTRI/R-CPI-NTR) — led it by ~16 months (Dec 2021: ALNRI +18% YoY while official CPI rent showed ~3%). The lag cuts both ways and was invoked whichever way fit the moment — leading data in 2022 to argue disinflation, the still-high lagged print in 2024 to justify higher-for-longer — the cleanest BLS-confirmed case of "doubt the headline." Full analysis.

Deeper — did money do it? (the causal question, honestly)

Proof of cause is outside what an exchange ratio can give, but the leading candidate — monetary expansion — can be probed as an association. Compare each multiple to M2 money-supply growth (Fed H.6, ×4.41 over 2000→2024):

The result cuts both simplistic stories. M2 grew ×4.41; consumer prices (×1.82) and wages (×2.02) rose under half as fast, while assets absorbed the gap and gold/silver exceeded it. That is exactly the footprint of monetary expansion showing up in assets rather than the shopping cart — strong association, and the most plausible mechanism. But it is not proof: no single money aggregate explains the cross-asset dispersion (gold ran ~2× M2, homes ~0.7×), and globalisation's goods-disinflation, interest rates, financialisation, EM central-bank gold buying and real-earnings growth are all uncontrolled confounders. "They printed money" is supported as a partial driver and refuted as a complete one.

Deeper — the mechanism: decompose each asset's rise

Proof of cause is beyond an exchange ratio, but each asset's rise splits — by price identity — into observable drivers, and that discriminates between the "why" stories far better than one M2 number. Equities: price = earnings × P/E. Housing: price = rent × price-to-rent. Gold: no cashflow, so its rise is monetary by elimination.

Equities rose on earnings, not monetary multiple — the P/E actually contracted; real profits roughly doubled (×2.3). Housing is mostly rents plus a rate/supply multiple. Gold alone is purely monetary. So the mono-causal stories both fail: "all money printing" cannot explain equities (multiples shrank), and "all fundamentals" cannot explain gold (it has none).

Which cause explains which asset

No row explains all three; no column is explained by one row. 'It's all money printing' fails equities (multiples contracted); 'it's all fundamentals' fails gold (it has none). The only cause that links WAGES to an asset gain is the labor->capital share shift, via equity earnings - documented accounting, not intent. The one cause that links wages to an asset gain is the labor→capital income-share shift: labor's share of GDP fell ~7% (-4.5 points to 2019), and the mirror is a rising profit share — the accounting bridge to the equity earnings that lifted the market. That is documented national-accounts arithmetic, not intent. Tier: the price identities are exact; the driver attributions are accounting decompositions plus association, not proof of cause; confounders (buybacks, globalisation, demographics, tax, foreign demand) are uncontrolled.

Deeper — what R₀ actually is: a conservation identity

One more level down, the ratio has a meaning that turns "the wage-hours went to owners" from rhetoric into an identity. The stock of an asset — shares of equity, ounces of gold above ground, housing units — is roughly fixed in the short run, and ownership shares of it sum to exactly 1. A price change creates no new units; it only revalues existing claims. One year of labor's real savings claims a fraction of that stock; at the higher price it claims R₀ of what it used to, and the complementary (1−R₀) is not destroyed — it stays with, i.e. transfers to, the existing holders.

So R₀ is not merely "labor bought less gold" — it is the shrunken share of a fixed pie that a year of work can pull from current owners, and the rest accrues to them. The transfer is definitional (a THEOREM, machine-checked: the unit-claim ratio equals the price-cross form exactly), who the owners are is empirical (the top-heavy ownership above), and why the price rose is unproven. Two honest caveats: real asset-unit growth (~1-2%/yr) is a small first-order correction, and the aggregate pie did grow — real US household net worth roughly doubled (×2.0). The claim is about labor's shrinking share-claim on that pie and its concentration, not "no wealth was created."

Falsify it yourself — the live certificate

Do not trust the numbers above; move them. Set your own wage growth, asset growth, and the tolerance you would grant the data, and the certificate recomputes: R₀ (labor's relative price), the breakdown error e*, the adversarial worst case, and the verdict. The claim is only as good as its ability to survive your inputs.

Where the proof stops — on purpose

The certified core is narrow by design. It does not establish:

• That the dollar was "debased." A fallen labor:gold ratio and a risen gold price are the same fact seen two ways; which one you name the mover is a modelling choice, not a theorem.
• That any actor, policy or institution caused it. This is correlation across series, not a mechanism.
• Any moral claim — "exploitation," "theft," intent. None follow from an exchange ratio.
• The exact magnitudes. Levels are representative and rounded; only the directions above are claimed, and only where e* clears 15%.

What remains, fully proven — stated as precisely as the evidence permits: over 2000→2024, a year of ordinary US labor came to command materially less of every liquid, independent store of value — gold, silver, equities — and against hard money that decline cannot be dismissed as a single-asset bubble (it holds across gold and silver), as data noise (it survives >26% simultaneous error in every input), or as a cherry-picked endpoint (it also certifies from a 2007 base). With authoritative primary series the certified basket widens to include homes on every price lens (median, constant-quality, per-square-foot). The sharper, CPI-dependent reading: real consumption wages roughly held; what collapsed was labor's claim on assets — the price of the store-of-value economy, measured in wage-hours, inflated several-fold — though the monthly mortgage carry is the one exception that stayed within noise, so the housing squeeze is about the down-payment, not the payment. On cause, monetary expansion is a supported partial driver, not a proven complete one. Every statement here carries its own certainty tier, and the honest boundary — long-horizon not last-decade, total not decomposition, price not carry, association not cause — is drawn exactly where each one fails.

The instrument hides the concentration

The proof above measures the typical worker. But the asset-price inflation that cost labor its wage-hours did not vanish — it accrued to whoever owned the assets. To see that transfer you have to look at wealth concentration, and here the most-cited government statistic is built so it cannot show you. This is not a claim that anyone fabricated a number; it is that the popular instrument — the Census money-income bracket table — has fine resolution where there is no concentration and none where all of it lives, while the Fed's own data shows the tail plainly. Judge "lie" versus "choice" yourself; the mechanism is demonstrable.

The open bucket — fine at the bottom, infinite at the top

The standard Census household-income table resolves the bottom in $10-15k steps, then dumps everyone from the comfortable to the billionaire into one open-ended "$200k and over" bin. No top 1%, 0.1% or 0.01% can exist in it — the concentration is quite literally off the chart, by construction.

What the Fed's own data shows

The Federal Reserve's Distributional Financial Accounts — also US government data — resolves the tail the income table erases. Net-worth share by group; the top 1% (dark) contains a top 0.1% of 13.8% on its own, and the bottom half of the country holds 2.5%.

Where the wage-hours went — who owns the inflated assets

This closes the loop with the proof. The gold, equities and housing that rose several-fold in wage-hours are owned overwhelmingly at the top (Survey of Consumer Finances). The purchasing power labor lost to asset prices is the wealth asset-holders gained — and Census money income, which excludes capital gains, never records it.

Why the headline gauge stays calm — the Gini hides the tail

Even the inequality number itself is built to under-react. On a synthetic US-shaped income distribution, doubling the income of the top 0.1% raises that group's share by +89% — but the headline Gini moves only from 0.462 to 0.49 (+6%, a drift it makes over a normal decade). A metric dominated by the middle cannot register a tail event; reported alone, it reads as stability during a concentration.

The structural distortions, itemised

Read the tags honestly: several are genuine distortions of what a reader thinks they are seeing; one is a plain statistical fact about the Gini; one is a confounder. Some choices have defensible motives (top-coding protects privacy; open bins protect small-sample reliability). The demonstrable point stands regardless of motive: the most-headlined instrument cannot show the concentration, and the government's own fuller series can. What you conclude about intent is your call — this page only draws the receipts.

The concentration, moving (1989 → 2024)

Not two snapshots — the trajectory. Federal Reserve net-worth shares over 35 years: the top 0.1% and top 1% climb while the bottom half flatlines near the floor (and briefly near zero after 2008). The same span, mirrored: a year of median labor fell from 100 to 18 in S&P-shares (1989=100). Concentration up, labor's asset-claim down, together.

The number is a choice — how each index sees the same tail

Take one distribution and double the income of the top 0.1%. Every inequality measure registers it differently, because each weights the tail differently. The Gini — the one almost always reported — moves least; top-share and Theil measures move most. Reporting a calm Gini during a tail event is not wrong arithmetic, it is a choice of instrument.

Atkinson at high inequality-aversion moves little too — but for the opposite reason: it looks at the bottom, not the top. The lesson is the same: no single scalar captures a distribution, and which one is headlined decides whether a concentration looks like a crisis or a calm.

Provenance & durability

Every number here is traceable. Of the 12 inputs, 9 are fetched live from the St. Louis Fed's keyless CSV mirror of the primary agencies (BLS, S&P CoreLogic, Census/HUD, Federal Reserve), refreshed by a committed script; the rest are documented snapshots (LBMA gold & silver, long-history S&P, S&P earnings) with citations. The figures reasoned from earlier reconcile with the live sources to within 4.8% — the hand estimates were right.

Cached to your browser — in case the data disappears

Government pages get revised, moved, and sometimes deleted. So on first view this page archives its full dataset to your browser's local storage, keeping the earliest copy you ever loaded alongside the latest. If the sources later change or vanish, your archived copy remains — and you can export it to a file.

Overlay, not proof: repricing strips monetary debasement out of a nominal figure, but it does not by itself establish cause. 2025–26 metal prices are annual-average approximations and provisional.