The Steel Mill the World Bank Said Should Not Exist

In 1966, South Korea produced less than 40% of the steel it consumed. Total consumption for the entire country came to about half a million tons a year, and there was not a single large integrated mill anywhere in it — the kind where ore goes in one end and steel coils come out the other.

Two years later, the Korean government decided to build one. And not a mill the size of the Korean market: a mill larger than the Korean market.

The international reaction was a rare consensus. Foreign governments and international agencies refused to take part in the project, judging it unviable at that scale. A World Bank study nailed down the logic: any mill larger than Korea's domestic market was doomed to fail. The financing never came. The mill was built anyway, in Pohang, and it was called POSCO.

If, at that moment, someone had been forced to bet on which emerging country would become a steel power, the obvious answer was not Korea. It was Brazil. Brazil already had three quality steel producers in operation, an incomparably larger domestic market, and iron ore at home. A decade later, the situation had reversed completely.

The 1985 scoreboard

In 1985, POSCO was running at 99% utilization of installed capacity. The largest Japanese steelmakers — the very ones that had taught the world how to make cheap steel — were averaging 52%. Part of that Japanese idle capacity was a direct consequence of POSCO's own entry into the world market: the newcomer was not sharing the pie, it was taking slices from the owners of the house.

That same year, the two large Brazilian state mills of the same industrial generation, Açominas and CST, were operating at a negative rate of profit: -3.5% and -0.83%.

The easy explanation is on the tip of everyone's tongue: management. The Korean state was competent, disciplined, cohesive; the Brazilian state was the Brazilian state. It is a comfortable story because it confirms what one already thought before looking at the numbers. The trouble is that when Nicolas Grinberg went to look at the numbers — in a London School of Economics doctoral thesis that reconstructs five decades of accounts for Brazil and Korea — the difference did not turn up where it was supposed to.

The math that doesn't add up

Start with the most elementary question: what it cost, in 1985, to produce one ton of cold-rolled steel coil at an efficient integrated mill in each country.

Read the last line again. Korea: 270 dollars. Brazil: 274. The gap is less than 2% — and Brazil also had the cheapest iron ore in the table, for the obvious reason that the ore was under its feet. From the standpoint of the cost of making steel, Brazil in 1985 was competitive. It is worth noting that these figures cover operating cost: they do not include administrative and financial expenses, nor, in some cases, depreciation. And it is precisely there, off the production line, that the difference was hidden.

It wasn't the steel. It was the mill.

POSCO had a simple, boring policy: buy equipment at the lowest price available anywhere in the world. The Brazilian state mills bought equipment and construction services from local firms, at substantially inflated prices. The result, measured as construction cost per thousand tons of installed capacity, is embarrassing:

• POSCO, in Pohang: $500

• Açominas: $1,000

• CST: $3,000

Açominas and CST were completed around 1985, within the same technological window. CST cost six times more, per unit of capacity, than the mill the World Bank had declared unviable.

And here comes the number that reorganizes the whole story. Grinberg redoes the accounts of the two Brazilian state firms, replacing the price paid for fixed capital with international prices — holding everything else equal, same operation, same workers, same market. The rates of profit stop being -3.5% and -0.83% and become 16.3% at Açominas and 6.8% at CST.

In other words: the Brazilian steelmakers were not bad companies producing expensive steel. They were viable companies that fell into the red because of the price they paid in order to exist.

Who was footing the bill, and why

This is not a story of incompetence, and this is where the thesis becomes interesting. Overpricing on domestic equipment is not an administrative accident: it is a transfer. Every extra dollar CST paid for a Brazilian part was a dollar that went into the till of a Brazilian capital goods maker that, at international prices, would have sold nothing at all.

Grinberg's central argument is that postwar Brazilian industrialization worked this way systematically — and could work this way because there was somewhere to draw from. The country generated extraordinary wealth that came not from industrial efficiency but from control of natural resources: farmland and ore sold at international prices. Economists call this ground rent. An overvalued exchange rate, tariffs, and subsidized public credit were the pipes through which part of that wealth drained from the primary sector into industry.

The effect is an industry that manages to be profitable producing at small scale, for the domestic market, without ever having to beat anyone abroad. Korea had no such cushion. There was no abundant ground rent to absorb a pricing mistake. Buying expensive equipment was not a bad policy choice: for Korean capital, it was simply unviable.

The cheap labor that wasn't so cheap

Go back to the 1985 cost table and look at the labor line: 63 dollars per ton in Japan, 25 in Korea, 26 in Brazil. Koreans and Brazilians tied. Except that the dollar lies.

When the Brazilian figure is corrected for real purchasing power — that is, when one asks what that wage actually bought in Brazil, instead of accepting what the exchange rate said — Brazilian unit labor cost rises from 26.10 to 41.89 dollars per ton. The Korean figure, under the same calculation, was 23.37. The Brazilian worker looked cheap because the exchange rate declared him cheap. The Korean worker was cheap.

How much this mattered to POSCO can be measured. Grinberg simulates what would have happened had the company paid Japanese wages in 1985: its rate of profit, calculated at export prices, would have fallen to around -2.3%. And if, on top of paying like Japan, its workers had been as productive as the Japanese? Profit would come back, but only to around 9%. The Korean advantage did not lie in superior productivity. It lay in labor cost — with inferior productivity.

The race nobody saw

There was also a quiet technical contest under way. Continuous casting is a technology that shapes liquid steel directly into semi-finished forms, skipping the old step of cooling it into ingots and reheating them. It eliminates an entire phase of production, saving energy and waste. In 1975, Korea and Brazil were both marginal users of the technique. After that, the paths diverge.

By 1990, Korea was already continuously casting 96.1% of its steel, having passed Japan. Brazil stood at 58.5% — below the United States, whose steel industry by then was the classic example of an industry in decline. Brazil would only reach the 90% range at the turn of the 2000s, when the race was over and everyone else was already there.

An honest asterisk

Two caveats keep this story from turning into a fable. The first: POSCO's profitability figures are calculated at export prices. At home, the company sold more cheaply — the literature the thesis engages with records those subsidized prices. POSCO, the champion of efficiency, was also a mechanism for cheapening inputs for the rest of Korean industry. It earned abroad and distributed at home.

The second: none of these profit figures is read straight off a balance sheet. They are estimates built on standardized assumptions about capital turnover and payment terms, applied the same way in both countries. That makes the comparison legitimate and the decimal precision somewhat less impressive than it looks.

The Brazilian epilogue

Between 1991 and 1993, Brazil's eight state-owned steelmakers — which accounted for roughly 85% of national steel output — were sold for a total of $8.2 billion, of which $2.6 billion was transferred debt. The average market discount on the value of the assets was 49%.

Half price. The mills that had cost twice and six times the world standard to build were handed over for half of what they were worth. Someone paid dearly to put them up, and someone else paid cheaply to take them away. The bill stayed where bills stay.

And the World Bank? The World Bank was not wrong. A mill that size was, in fact, too big for the Korean market — the arithmetic was correct. What the report did not consider was the simplest hypothesis of all: that Korea had no intention of selling to Korea.