Supply, demand and the cost curve: Demand: derived, cyclical and tied to development

1.1 Demand is derived, so buyers barely notice the price

Copper demand can be measured at seven points along the chain, Tilton and Guzmán show, from smelters buying concentrate to households buying finished goods. Statistics usually exist only for fabricators, so that is what analysts call demand, knowing it is a stand-in. The demand that sets the price is market demand: every buyer, including stockholders and speculators.

Because the material is wanted only for what it makes, two things follow. First, its price is usually a small share of the final product's cost. Doubling or tripling the price of lead or platinum adds less than 1 per cent to the cost of making a car. Radetzki and Wårell make the same point for wheat in bread and lead in batteries. A buyer whose own selling price hardly moves has little reason to cut purchases, so the short-run price elasticity of demand is low.

Second, changes in final demand are amplified on the way back up the chain. Each stage holds stocks in proportion to its sales. When sales rise, each stage buys for the extra sales and to top up its stocks, passing a bigger order to the stage before. The raw material producer at the top sees the largest swing. This is one reason metal demand is more volatile than the output of the goods that contain it, and why destocking can make a mild recession feel like a collapse at the mine.

So ask where in the chain a demand figure was measured, and whether it includes stock changes, which can dominate a single year.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 30-35
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf pp. 110-111

1.2 Intensity of use rises and then falls as countries develop

Tilton and Guzmán start from an identity. Consumption of a metal equals GDP times intensity of use, and intensity of use sums, over every product, the metal per unit of product times that product's share of output. So demand changes because the economy grows, the mix of goods changes, or the metal content of each good changes.

The intensity-of-use hypothesis adds a claim about the mix. A poor farming economy uses little metal per dollar. Industrialisation brings roads, power lines, factories and flats, so intensity rises. Once those are built, spending moves to services and intensity falls. Against income per head, intensity traces an inverted U; Radetzki and Wårell show it for steel in South Korea from 1970 to 2017. On the rising limb demand grows faster than income, so the long-run income elasticity is above 1; on the falling limb it is below 1. So a 1 per cent rise in Chinese GDP moves metal markets more than the same rise in American GDP.

The curve is a weak forecasting tool. Prices and technology change the metal content of products over time, independently of income. Each year's observation sits on a different true curve, usually a lower one as products get lighter, so a curve fitted to history matches none of them and fails exactly when the past stops being a guide. Humphreys records the drift: world intensity of use fell 25 per cent for nickel and 47 per cent for steel between 1970 and 2000, as copper tube walls thinned, cars shrank and zinc coatings got thinner.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 44-48
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf p. 22
• Humphreys (2015), The Remaking of the Mining Industry · pdf pp. 26-27

1.3 China's urbanisation produced the biggest demand surge in history

In 2000, executives at a large iron ore company tried to guess China's steel output ten years ahead. China then made about 125 million tonnes; the meeting settled on 180 million, 200 at most. In 2010 China made 639 million tonnes.

Humphreys traces the surge to three linked drivers. Exports, boosted by entry to the WTO at the end of 2001, made China the workshop of the world. Investment rose past 40 per cent of GDP in 2003 and to 48 per cent by 2010. Urbanisation took the urban share from about 18 per cent in the late 1970s to 30 per cent in 1995 and 50 per cent in 2013, putting 345 million more people in cities than in 1995. An urban household uses an estimated 10 to 15 times as much steel as a rural one, and scarce farmland pushed cities upward: a 35 to 50 storey block needs about twice the steel per square metre of an 8 to 12 storey one.

The result was the most material-hungry growth on record. China used nine times as much steel per million dollars of GDP as the US. Its share of world copper use rose from 11 per cent in 1998 to 47 per cent in 2013, and of steel from 16 to 47 per cent. Radetzki and Wårell put China's share of world copper demand growth in 2005 to 2016 at 124 per cent, so use fell elsewhere. China added 6.5 billion tonnes of steel to its in-use stock in 2006 to 2015 (Smil).

• Humphreys (2015), The Remaking of the Mining Industry · pdf pp. 51-73
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf p. 136
• Smil (2016), Still the Iron Age · pdf pp. 203, 207, 230

1.4 Income moves demand fast; price moves it slowly

Two elasticities summarise demand: the percentage change in demand for a 1 per cent rise in income, and for a 1 per cent rise in price. Their short- and long-run values differ in opposite directions.

In the short run, which Tilton and Guzmán put at up to about five years, income dominates. Most metal goes into construction, capital equipment, transport and consumer durables. These boom harder than GDP in upswings and slump harder in recessions, so the short-run income elasticity is usually above 1.

Price works the other way. In the short run demand barely responds, because the metal is a small cost share and switching materials needs new equipment, retraining and requalified products. Over several years those changes get made. The conventional presumption Tilton and Guzmán report is a price elasticity below 1 in the short run and above 1 in the long run, so short-run demand curves are steep and long-run ones flat. High prices do cure high prices, but late, and the cure tends to stay, because a firm that has paid to switch materials rarely switches back. In 2010 a spokesman for US copper and brass fabricators told regulators that substitution was the 'even more damaging' effect of copper's price doubling (Humphreys).

Food breaks the pattern. Under Engel's law households spend a falling share of income on staples, so their income elasticity is low, and food demand varies little over the business cycle. Grain price swings usually start on the supply side.

• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 35-36, 41-44, 62-63
• Radetzki & Wårell (2021), A Handbook of Primary Commodities · pdf p. 41
• Humphreys (2015), The Remaking of the Mining Industry · pdf p. 115

See it in the table

Check your understanding

Final demand for a good rises 5 per cent. Three downstream stages each keep stocks equal to half a period's sales. Roughly how much do orders for the raw material rise?

Why does a curve fitted to past intensity-of-use data tend to mislead forecasters in a rich country?

In the short run, what are the typical values of the income and price elasticities of metal demand?