What is an ounce of gold worth in real assets?
Since gold began to float in 1971, its price has tracked scarce real assets like farmland and copper. Two models, cows + farmland and farmland + copper, set a fair-value band from 50 years of data. The gap between that band and today's price is the monetary premium investors pay for gold.
Loading the 100-year series…
A century of gold vs. its real-economy value
Annual reconstruction, 1926–2025, plus the 2026 reference point. Hover or tap for any year.
The blue band spans the two hard-asset fair values (from 1960, when the copper data starts). The gray area marks the Bretton Woods era, when the U.S. fixed gold at $35, so the market price could not respond to the real economy. The float began in 1971. Log scale shows percentage moves at equal size across a century of inflation.
Where today sits in history
The premium only means something compared with how gold has usually traded against the same yardstick.
Bands: deeply cheap < −20% · real-value zone −20% to +15% · monetary premium +15% to +40% · high premium +40% to +80% · overshoot > +80%.
What anchors fair value
Change the inputs. Everything on the page recomputes.
The index is the geometric mean of one cow and 0.6 acres. Fair value scales it by a fitted constant and slope: since 1972, gold has risen about 1.5% for every 1% rise in the index. Over 300 years, though, an ounce has bought roughly the same number of cows: about 1.3 in the 1700s and 1.1 today, which implies a long-run slope near 1. The 1.49 slope is specific to the floating-dollar era since 1971. The farmland + copper model has no slope to fit: one ounce is worth a fixed bundle of the two (their geometric mean), with only the level calibrated to 1972–2025.
Two models, one band
Neither is the "true" value of gold. The shaded band on the chart spans the two, so each model checks the other: gold above both is a stronger signal than gold above one. Their caveats differ.
Cow + farmland
- The slope is mostly about cheap cows. Cow prices fell about 65% after inflation from 1980 to 2020 while milk per cow doubled. Farmland alone tracks gold with a slope near 1.0, so the 1.49 slope mainly reflects modern dairy farming making cows cheaper, not gold.
- It doesn't hold over centuries. An ounce bought about 1.3 cows in the 1700s and about 1.1 today, a long-run slope near 1. Extending 1.49 forward assumes the post-1971 pattern continues.
- One more fitted number. A fitted slope can describe history well and still drift. Set the slider to 1.00 to see the level-only version.
- Thin early data. Some cow-price years are interpolated.
Farmland + copper
- Chosen after testing about 30 benchmarks. Scarce assets (land, copper, silver, oil, homes) tracked gold one-for-one; farmed goods didn't. Picking the best of many tests flatters the winner, so treat its error as optimistic.
- Copper is cyclical. Chinese construction, grid and EV buildouts, and mine disruptions can move it sharply. Copper's 2026 surge lifted this fair value by about a fifth in a year.
- Shorter history. Copper prices here start in 1960, so the line only covers the last two-thirds of the chart.
- Gold has done the adjusting, so far. When gold got expensive relative to copper or farmland, gold fell back over the next one to five years while they barely moved. Sustained central-bank buying could change that pattern.
Both Fit on about 54 annual observations since gold began to float in 1971, with the same data used to fit and score them except for the out-of-sample test. They describe where gold has tended to trade, not where it will go, and are useless for short-term timing.
12-month scenarios
Uses the cow + farmland model. Fair value grows at its trailing 10-year rate. Each scenario sets the premium a year from now. The mean-reversion case uses the speed at which the premium has historically closed.
| Scenario | Probability | Premium in 12 months | Implied gold | vs. today |
|---|---|---|---|---|
| Historical mean reversion— | % | — | — | — |
| Premium holdsToday's premium persists | % | — | — | — |
| Monetary stressYour assumption | % | % | — | — |
| Severe stressYour assumption · Jan 2026 peak was ≈ +90% | % | % | — | — |
| Probability-weighted | 100% | — | — |
How good is the cow + farmland model?
Computed live from the published data. Any two prices that both rise with inflation will look correlated, so the tests below strip out that shared trend.
Looks strong, but mostly reflects a century of shared inflation.
The floating-era fit on a log scale, with fitted constant and slope.
Do gold and fair value move together in the same year? Barely.
How long it has historically taken half of a valuation gap to close.
In-sample All statistics are fit and measured on the same data, so they describe history rather than prove predictive power. The model is a slow-moving yardstick, not a timing tool.
How good is the farmland + copper model?
The same tests, computed live. This model has no fitted slope, only a level, so there's less room to flatter the fit.
Strong, but like any price pair it partly reflects shared inflation.
The floating-era fit on a log scale, with the slope fixed at 1.
Moves with gold within a year more than cows and farmland do, but still loosely.
How long it has historically taken half of a valuation gap to close.
In-sample Measured on the same data used to set the level, except the out-of-sample test (fit on 1972–2000, typical error on 2001–2025: —). Copper was picked after testing about 30 benchmarks, so these scores are somewhat optimistic. Copper prices here start in 1960.
Do the models beat plain inflation?
The simplest rival yardstick is the consumer price index, fitted to gold the same way. If a hard-asset model can't beat that, it adds nothing. Floating era, 1972–2025; the best score in each row is ticked.
| Test | Cow + farmland | Farmland + copper | Inflation only (CPI) |
|---|
Ounces of gold per real asset
How many ounces it took to buy one cow, one acre, or one tonne of copper. Lower means gold bought more.
Per productive cow
Per acre of U.S. farmland
Per tonne of copper
Before 1867 the cow series is sparse: decadal New England cattle prices (1720s–1770s) converted at the British post-1717 mint equivalent, a 1790s New York benchmark and an 1825 Detroit probate benchmark. These show direction, not a seamless continuation of the USDA series.
Method and sources
Fair value
fair_value = ea × indexb
a, b = least squares on ln(gold) vs ln(index), 1972–2025
Farmland + copper
k = average of ln(gold ÷ √(land × copper)), 1972–2025
Premium and scenarios
fair_12m = fair_value × (1 + g₁₀)
price_12m = fair_12m × (1 + premium_12m)
reversion: ln(1+p)₁₂ = ln(1+p) × (1 + β)
g₁₀ is the trailing 10-year growth rate of fair value. β is the slope from regressing each year's change in log premium on the prior year's level (1972–2025).
Data
- Gold: annual U.S. dollar price per troy ounce; official/fixed price before 1968, London market thereafter.
- Cows: USDA NASS price of milk/replacement cows, $/head, with USDA ERS annual dairy-cow prices filling 1981–2019; remaining gaps (mostly 1938–1979) linearly interpolated.
- Inflation: BLS CPI-U, U.S. city average, annual average (FRED CPIAUCNS). 2026 is estimated by carrying forward 2025's inflation rate.
- Farmland: USDA farm real-estate value, $/acre (Census of Agriculture and Land Values surveys).
- Copper: LME copper, $/metric ton, annual averages from the World Bank Pink Sheet (1960–2024) and IMF via FRED (PCOPPUSDM) for 2025 and the latest month.
- 2026 reference values: gold $3,628 (Nov 19, 2026), cow $3,160 (Oct 2026), farmland $4,500 (2026), copper $14,473 (Sep 2026). Provisional.
Why cows and farmland?
Both are productive, scarce, real assets that reprice freely and have long U.S. price records. Ancient and medieval sources often value a cow somewhere between a fraction of an ounce and about an ounce of gold, and modern U.S. replacement cows have returned to roughly that range. Farmland offers the cleanest long-run scarce-asset comparison: gold's purchasing power over land collapsed while gold was pinned at $35, then recovered after the float.
Why add copper?
We tested 22 benchmarks against gold from 1972 to 2024, fitting each on 1972–2000 and checking it on 2001–2024. Scarce, hard-to-produce assets (farmland, copper, silver, oil, homes) tracked gold roughly one-for-one. Farmed and manufactured goods (wheat, corn, beef, cows, cotton, aluminum) got cheaper over time as production improved, so gold outran them. A simple farmland + copper bundle with no fitted slope had among the lowest out-of-sample errors, and when gold strayed from it, gold was the one that moved back. Manufacturing wages were also tested and added nothing beyond cows and farmland, so labor was removed.
How is fair value calibrated?
The model originally used fixed anchors: one cow or 0.6 acres per ounce. Against those, gold traded about 45% below "fair value" in a typical floating-era year. Shifting the line down fixed the average but left today's fair value near $1,600, because gold has outgrown cows and farmland over time. So fair value now has a fitted slope as well as a level: since 1972, gold has risen about 1.5% for every 1% rise in the cow-and-farmland index. That fits better (log R² 0.84 vs. 0.75) and, fitted on 1972–2000 alone, tracked 2001–2025 more closely than the level-only version. It also means "cows per ounce" drifts upward over time instead of staying fixed. A slope is one more fitted number, so treat it as a description of history. Before 1971 the gold price was fixed, so there's nothing to fit, but long-run purchasing power is a check: an ounce bought about 1.3 cows in the 1700s (colonial New England, 1725–1796) and about 1.1 today, which implies a slope near 1 over 300 years. The 1.49 slope is specific to the floating-dollar era since 1971. The CSV keeps the original anchors in modeled_gold_usd_oz; the page fits the slope on top. Set the slope slider to 1.00 to see the level-only version.
What changed in this version?
Added a second fair value, farmland + copper, with a three-way comparison against inflation and caveats for both models. Removed the labor diagnostic. Fair value now uses a fitted level and slope against the cow-and-farmland index (1972–2025), with an out-of-sample check. Connected live gold and copper data. Replaced 36 interpolated cow-price years (1981–2019) with USDA ERS annual dairy-cow prices. Added an inflation-only (CPI) baseline to the chart and a head-to-head test against it. Corrected the 1931 and 1949 gold prices to the official prices of $20.67 and $34.72. Added log scale, a floating-era view, historical percentile, live model diagnostics, fair-value growth in the 12-month scenarios and a mean-reversion case calibrated to history.