Supply, demand and the cost curve: Substitution, technology, and demand for food, feed and fuel

2.1 Substitution is driven more by technology than by price

The traditional view treats substitution as cost-minimising. Tilton and Guzmán's example is a firm making 60,000 light trucks from steel and aluminium. Many combinations will do; the curve linking them is an isoquant. With aluminium at $2,500 a tonne and steel at $1,000, $50 million buys 20,000 tonnes of aluminium, 50,000 tonnes of steel, or any mix between. The firm picks the mix where that budget line just touches the isoquant. If steel doubles to $2,000, the $50 million buys only 25,000 tonnes of steel, the line tilts and the cheapest mix moves toward aluminium. Fewer trucks may also be sold, so aluminium demand can rise or fall.

The alternative view says substitution mostly does not happen this way. Real isoquants are often single points, a few discrete processes or straight lines, so prices can move a long way without changing the mix at all. What changes the mix is new technology that shifts the isoquant itself: new alloys, polymers, composites and processes. Regulation can also force a switch, with no help from prices, as when the US banned lead paint in homes in 1978 and paint makers moved to titanium dioxide.

Price and technology interact, because high prices fund the research that produces substitutes. Counting this indirect effect, Tilton and Guzmán argue, the long-run price elasticity is much larger than direct substitution suggests, and the demand curve is no longer reversible. A spike that triggers a new technology lowers demand for good, even after the price comes back down.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 48-56

2.2 Today's switches: aluminium for copper, iron phosphate for cobalt

Copper is the best conductor in wide use, but not the only one. Conway notes that aluminium conducts well enough and is much lighter, so it is often used for long-distance high-voltage lines strung overhead. Radetzki and Wårell use the pair to show how substitution limits market power: a copper cartel acting alone would push buyers toward aluminium.

Batteries show technology creating and reshaping demand. Batteries took 88 per cent of lithium use in 2025, according to USGS. The cathode recipe decides which other metals a cell needs. Conway lists lithium cobalt oxide in phones, nickel-manganese-cobalt (NMC) in most electric cars, and lithium iron phosphate (LFP), which is more stable but stores less energy; Scheyder notes LFP tends to be cheaper. Conway's typical car battery holds about 40 kg of lithium, 10 kg of cobalt, 10 kg of manganese and 40 kg of nickel. A shift to LFP moves demand from nickel and cobalt toward iron and phosphate without changing the number of cars. Concern over child labour in Congolese cobalt mining led Elon Musk to promise in 2018 that future Teslas would use no cobalt, a goal not yet met when Scheyder wrote. Cobalt shows the mismatch this creates. Battery makers can move its demand quickly, but Nassar, Graedel and Harper class cobalt among the metals obtained mostly as companions, chiefly of copper and nickel.

Technology destroys uses too. If electric cars replace combustion engines, Tilton and Guzmán note, demand for platinum and palladium in catalytic converters, their main market, would collapse.

• Conway (2023), Material World · pdf pp. 197, 311-312
• Scheyder (2024), The War Below · pdf pp. 20, 239
• USGS Mineral Commodity Summaries 2026, Lithium web ↗ · Domestic Production and Use
• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf p. 36
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 219-220
• Nassar, Graedel & Harper (2015), Science Advances web ↗ · pdf p. 2

2.3 Food, feed and fuel: demand set by diets and policy

Farm demand differs from metal demand in three ways. Staple food has a low income elasticity, as Engel's law predicts, and barely moves with the business cycle. And much grain is not eaten by people at all but fed to animals or turned into fuel, so its demand derives from meat consumption and energy policy.

Feed links grain to diets. Richer people eat more meat, which takes several kilograms of grain per kilogram. Morgan describes Soviet planners raising feed grain use by 40 per cent between 1968 and 1971 to build up herds. Even after bumper harvests in 1970 and 1971 their traders bought 7.8 million tons of grain, seeking corn for livestock rather than wheat for bread.

Fuel links grain to oil. Geman reports that by the early 2000s over 5 per cent of US corn went to ethanol, and about 95 per cent of North American ethanol came from corn. USDA's October 2026 projection has ethanol taking 5,650 million bushels of US corn in 2026/27, about 35 per cent of total use of 16,305 million and close to feed at 6,000 million. Radetzki and Wårell put world biofuel output at 1.9 million barrels a day in 2018. Blending mandates make this demand insensitive to grain prices. Baffes and Haniotis, cited by Radetzki and Wårell, found that the stronger link between energy and other commodity prices was a dominant cause of the 2006 to 2008 food price spike; biofuels alone, grown on about 1.5 per cent of grain and oilseed land, were not.

• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 41, 85, 138
• Morgan (1979), Merchants of Grain · pdf pp. 207, 213, 231
• Geman (2005), Commodities and Commodity Derivatives · pdf pp. 160-161
• USDA WASDE, October 2026 web ↗ · U.S. feed grains and corn supply and use

See it in the table

Check your understanding

According to Demler's study of US beer containers from 1950 to 1977, what mainly drove the shifts in material market shares?

Why does a carmaker's move from NMC to LFP cathodes matter for metal markets even if car sales are unchanged?

In USDA's October 2026 projection for 2026/27, roughly what share of US corn use goes to ethanol?