Supply, demand and the cost curve: Inelastic supply: spikes, lead times, by-products and harvests

4.1 Steep demand meets vertical supply, and prices spike

Radetzki and Wårell draw the short-run supply curve as existing plants ranked by variable cost. With spare capacity it is flat, and extra demand is met with little change in price. Near full use it steepens; at capacity it is vertical. Tilton and Guzmán add the other two ingredients of volatility: short-run demand is price-inelastic, and it shifts a lot over the business cycle. A steep demand curve swinging back and forth against a supply curve that ends in a wall turns small shifts in demand into large moves in price.

Copper averaged $1,559 a tonne in 2002 and $8,828 in 2011; nickel went from $5,293 in 1993 to $37,230 in 2007 (Radetzki and Wårell). Humphreys records copper more than doubling between September 2005 and May 2006, from $3,857 to $8,044 a tonne, with the spot price over $2,000 above the 27-month price. That backwardation was the market paying a premium for metal now.

Was it speculation? Studies by the IMF, the UK Treasury and others found little evidence that fund money set price levels, and iron ore and minor metals, with no futures markets, also spiked. Humphreys' verdict: demand shifting right against vertical supply, with the price rising until it rationed the metal.

Minerals and crops differ in where shocks start. Metal volatility comes mainly from demand, so price and output rise and fall together and miners' revenues swing violently. Crop volatility comes mainly from weather, so a short harvest raises the price and partly cushions farm revenue.

• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 102-103, 110-111
• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 112-114
• Humphreys (2015), The Remaking of the Mining Industry · pdf pp. 105, 119-120
• Yergin (1991), The Prize · pdf pp. 1049, 1062-1063, 1081

4.2 Long lead times turn price signals into investment cycles

New mineral supply passes through four stages: discovering resources, proving reserves, building capacity and producing. Estimates of the lag differ by author. Tilton and Guzmán say a new mine typically takes about five years, sometimes up to a decade. Radetzki and Wårell put the lag from exploration to production at five to ten years, sometimes more. Humphreys says a large mine can take eight to ten years.

The lags create a cycle. Prices rise, supply cannot respond, and prices keep rising. Projects compete for the same engineers and equipment, so capital costs soar. A tonne of annual copper capacity, long reckoned at about $5,000 to build, cost $10,000 to $15,000 on many large projects by 2010; greenfield iron ore capacity went from about $100 to $200 a tonne. Mining capital spending peaked in 2011 and 2012, about a year after prices.

Then the capacity arrives together. Humphreys shows that the copper price upturn of 1987 did not accelerate mine output until 1995. On the same lag the 2004 upturn should have shown up in 2012, and it did: output, which had grown 1.7 per cent a year over 2001 to 2011, rose 4 per cent in 2012 and 8.5 per cent in 2013. The largest miners' returns on capital fell from 22 to 23 per cent at the peaks to 5 per cent in 2013.

What the boom meant is disputed. Radetzki and Wårell reject the super-cycle: the boom was a long disequilibrium, booms breed false belief in a higher normal price and so overinvestment, and China's rise is unlikely to lift equilibrium prices. Humphreys doubts it was a cycle, but argues the old downtrend in real prices will not soon return, because costs and the dollar may have moved prices to a higher plane.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 67-68, 188
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 42, 130, 137-139
• Humphreys (2015), The Remaking of the Mining Industry · pdf pp. 106, 110, 114, 146-150, 228
• USGS Mineral Commodity Summaries 2026, Lithium web ↗ · Salient Statistics; Events, Trends, and Issues

4.3 By-products are supplied whatever their own price does

Many ores yield more than one product. Tilton and Guzmán define the roles by what drives the amount of ore mined. A main product's price alone decides it. Co-products share that role, as lead and zinc often do. A by-product's price has no influence on ore output at all; if it did, it would not be a by-product.

Two features follow. First, by-product supply hits a ceiling: you cannot recover more molybdenum than the copper ore contains, so near that limit supply turns vertical, even in the long run. The ceiling moves only when the host's market moves. Second, only costs specific to the by-product count, since the host bears the joint costs of mining and milling. By-product supply can be cheap but not free: if the price stays below the specific costs of separation and refining, the recovery circuit will not be replaced.

Nassar, Graedel and Harper call the share of a metal's primary output that comes as a companion of a host its companionality. Of 62 metals, 38 exceed 50 per cent. Indium sits in zinc ore at a few parts per million, too little ever to mine for itself. At least 90 per cent of selenium and tellurium is tied to copper, and about 71 per cent of silver comes from other metals' ores. The shares move: molybdenum's companionality fell from about 75 to 47 per cent over 1985 to 2013 as dedicated mines gained, while silver's stayed at 70 to 77 per cent. Eighteen companions, including germanium, terbium and dysprosium, also have concentrated production and almost no recycling. In the table, Gallium, recovered from bauxite processing, shows China with about 99 per cent of 2024 output.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 65-66, 76-77, 93
• Nassar, Graedel & Harper (2015), Science Advances web ↗ · pdf pp. 1-3
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 105-106

4.4 Farm supply answers within a season, but weather rules and stocks buffer

Annual crops are the exception to long lead times. Radetzki and Wårell note that wheat, sugar or bananas can be expanded between seasons by planting more land, while coffee, palm oil and other tree crops take years, like mines. That fast response is one reason they rule out cartels in cereals and sugar.

Within a season, supply is set by what was planted and by the weather. Yields swing, and so does harvested area when drought makes fields not worth harvesting. USDA's October 2026 projection for US wheat in 2026/27 has harvested area falling from 37.3 to 31.9 million acres, yield from 53.3 to 48.1 bushels an acre and output from 1,989 to 1,534 million bushels; the average farm price rises from $5.06 to $6.30 a bushel. The Wheat tile shows US hard red winter wheat at $347 a tonne in September 2026, against $233.8 a year earlier.

Stocks are the buffer. Geman sets out the grain balance sheet: supply is carry-in plus production plus imports; demand is domestic use plus exports; the remainder is carry-over, which links one crop year to the next. The key ratio is ending stocks divided by use. Her example, 540 million bushels against 2.7 billion of use, gives 20 per cent, and she treats 20 to 40 per cent as the usual range. When the ratio is low, any supply shock sends prices very high. For the world, USDA projects 2026/27 wheat ending stocks of 276 million tonnes against use of 828 million, about a third. Storability makes supply more elastic than production alone, which is why metal traders watch exchange and port stocks just as closely.

• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 41-42, 223-224
• Geman (2005), Commodities and Commodity Derivatives · pdf pp. 154-155
• USDA WASDE, October 2026 web ↗ · U.S. wheat supply and use; world and U.S. supply and use for grains
• Morgan (1979), Merchants of Grain · pdf pp. 219-222, 228, 231

See it in the table

Check your understanding

Yergin puts the December 1973 net oil supply loss at about 9 per cent of non-communist supply, yet posted prices roughly quadrupled. What does this imply?

Why does by-product supply become price-inelastic even in the long run?

A grain market has ending stocks of 540 million bushels and annual use of 2.7 billion bushels. What is the stocks-to-use ratio?