Why does gold behave like no other commodity?
Every mechanism in this course — inventories, elasticity, the capex cycle, industrial demand — describes a commodity that gets consumed. Gold is not consumed. Nearly all the gold ever mined still exists, which breaks the model completely and turns gold into a portfolio asset wearing a commodity's clothes.
The stock-to-flow fact that changes everything
Above-ground gold stocks are on the order of 210,000 tonnes. Annual mine production is roughly 3,600 tonnes, with about another 1,200-1,300 tonnes from recycling.
- Annual mine supply ÷ existing stock ≈ 1.7%.
Compare oil, where essentially a whole year's production is burned within the year, so the flow balance is the market. For gold, a spectacular mine discovery or a major mine outage changes the available stock by a fraction of a percent. The flow balance barely matters.
What sets the price instead is whether the owners of the existing 210,000 tonnes are willing to keep holding it, and at what price newcomers want some. Gold's price is set by portfolio preference, not by physical scarcity this quarter — which is why it responds to interest rates and monetary conditions rather than to inventories and industrial cycles.
Three widely studied influences
Real interest rates. Gold pays no coupon, no dividend and no rent, and costs something to store and insure. Its opportunity cost is therefore the real yield available on inflation-linked government bonds: hold gold, forgo that yield. Historically this has produced a negative relationship — higher real yields associated with a weaker gold price. The honest caveat is that the relationship visibly weakened in 2022-2024, when real yields rose sharply and gold did not fall as the framework implied. A model that stops working is information, not something to explain away.
Monetary and debasement narratives. Gold has no issuer and no counterparty, which is exactly why some buyers hold it against currency, sovereign or institutional risk. That demand is narrative-driven and essentially unmodellable — a genuine analytical limitation to state rather than to paper over.
Central bank demand. Official-sector net purchases exceeded 1,000 tonnes in each of 2022, 2023 and 2024 — roughly a fifth of total annual supply, from a buyer that is largely price-insensitive and policy-driven. Note the direction can invert: through the 1990s the official sector was a persistent net seller. Same asset, opposite official flow, thirty years apart.
The record, stated without flattery
Gold's reputation as an inflation hedge does not survive medium-horizon testing. From its January 1980 peak, gold's nominal price was still far below that level two decades later — down roughly two-thirds — while US consumer prices roughly doubled over the same period. A buyer at that peak had lost something on the order of 85% of their purchasing power twenty years later, in the asset most often described as a store of value.
Over multi-century horizons the picture is kinder, and long-run studies find gold's real price wandering around a level rather than collapsing. Both facts are true simultaneously. The relevant question is which horizon a real investor lives on: twenty years is a substantial fraction of an investing life, and "it works over centuries" is not a property anyone can use.
In the data
Gold's opportunity cost is published every business day: the real yield on inflation-linked Treasuries, what a safe dollar asset pays above inflation. The latest ten-year figure is below.
On 28 September 2026 it stood at 2.90%. That is what a holder of gold, which pays nothing, gave up that day. The real yield and the gold price keep different calendars, US Treasury business days against futures trading days, so pair them date by date rather than row by row.
Try it now
- The gold history on this page begins in August 2000, so it cannot show you the 1980–2000 stretch above. The window it does hold is below at its full length, with US inflation, one figure a year, under it. Run the purchasing-power arithmetic on 2001 to 2025, and note that any claim about the 1980s needs a source this page does not contain.
- Set gold against a real-yield series over the last fifteen years, and find the stretch where the usual relationship broke down. The table below is the last close of each year for continuous gold futures beside the ten-year real yield on the same day, measured 28 September 2026. The framework says the two columns should move in opposite directions; mark the years where they did and the years where they did not.
| year end | gold, $ per troy ounce | 10-year real yield, % |
|---|---|---|
| 2011 | 1,565.8 | -0.07 |
| 2012 | 1,674.8 | -0.67 |
| 2013 | 1,201.9 | 0.80 |
| 2014 | 1,183.9 | 0.49 |
| 2015 | 1,060.3 | 0.73 |
| 2016 | 1,150.0 | 0.50 |
| 2017 | 1,306.3 | 0.44 |
| 2018 | 1,278.3 | 0.98 |
| 2019 | 1,519.5 | 0.15 |
| 2020 | 1,893.1 | -1.06 |
| 2021 | 1,827.5 | -1.04 |
| 2022 | 1,819.7 | 1.58 |
| 2023 | 2,062.4 | 1.72 |
| 2024 | 2,629.2 | 2.24 |
| 2025 | 4,325.6 | 1.93 |
- Write one sentence distinguishing gold from copper in terms of what sets each price. If your sentence mentions inventories for gold, try again.
A note on what we do here. EODHD Academy teaches how markets work. This lesson describes what has historically driven gold's price and what the evidence does and does not support. It is not a recommendation to hold gold, avoid gold, or expect any particular price, and it is not a substitute for advice from a qualified professional.