Data Desk

The Input That Does Not Set the Price: Iron Ore, Steel, and a 0.14 Correlation

Steel buyers keep hearing that iron ore justifies the next price increase. Thirty-four years of monthly data say the two barely speak, and that should change how you read the surcharge letter.

The surcharge letter usually arrives with a chart of iron ore attached. The input is up, the supplier explains, so the price of steel must follow, and everyone shares the pain. It is a tidy argument with one problem: the archive has been keeping score, and the score is 0.14. That is the correlation between the monthly percent changes of the global iron ore benchmark and the monthly percent changes of the Bureau of Labor Statistics producer price index for iron and steel, measured over the 415 overlapping months from January 1992 through July 2026. Ore trades at $102/tonne as of Jul 2026, up 0.4% from a year earlier. What the domestic steel index does next has remarkably little to do with it.

A 2.0% explanation

Correlation invites hand-waving, so square it. A correlation of 0.14 means iron ore's month-to-month moves account for about 2.0% of the variance in the steel index's month-to-month moves. The other 98.0% comes from everywhere else: scrap flowing into electric furnaces, energy, labor, mill utilization, freight, trade policy, and the pricing discipline of a consolidated domestic industry. Nor is this the quirk of a quiet stretch of history. The window opens in January 1992 and runs through July 2026, and inside it the benchmark traveled from $11 in January 1994 to $216 in June 2021, a 18.8x span. The input made one of the wildest journeys in the commodity archive, and its co-movement with the American mill price still rounds to noise.

Iron ore benchmark, Jul 2026: $102/tonne. The archived range runs from $11 in January 1994 to $216 in June 2021. Across that entire span, the correlation with the domestic steel index's monthly moves is 0.14.

The last five years make the same point without statistics. From year-end 2021 to the latest reading, ore fell 8.0%, from $110.46 to $101.60 a tonne, and it now sits at the 44th percentile of its archive, 52.9% below its June 2021 peak of $215.82. A mid-range input, in other words. The iron and steel producer price index, meanwhile, prints 374.10index (1982=100) as of Jul 2026, which its archived history places at the 82nd percentile. The input drifted back toward its middle; the output stayed parked near the top of its range. If ore set the price, those two readings could not coexist. They coexist comfortably, because the American steel invoice is mostly not iron ore.

What actually sets the American steel price

The disconnect is structural, not cyclical. Most American steel is melted from scrap in electric arc furnaces, a route that does not buy iron ore at all, and scrap has its own market, tied to demolition, collection economics, and export demand rather than the seaborne ore trade. On top of the metallurgy sits policy: tariffs and quotas insulate the domestic price from the global one, so a Pacific ore rally can wash up against the border and stop there. And mills price to order books and capacity, not to a cost-plus formula. When demand is strong, the index climbs whether ore cooperates or not; when it thins, discounts appear even as ore firms. The input matters to a mill's margin. It does not set the market's price.

One honest caveat, because the methodology deserves to be argued with. A correlation of monthly percent changes measures co-movement at monthly speed; it does not rule out slower connections. Over a span of years, a sustained ore regime can seep into blast furnace economics and, through them, into the index. But surcharge letters do not cite decade-long regimes. They cite last quarter's chart, and at that frequency the connection is the 2.0% kind, not the kind that justifies repricing an order book.

If your steel price is indexed to iron ore, it is tied to a number that explains almost none of the invoice it claims to predict.

The escalator that points the wrong way

Put dollars on it. Take a fabricator buying $750,000 of steel a year who accepted an ore-indexed escalator at year-end 2021, the kind of clause that sounds rigorous across a negotiating table. Ore then fell 8.0% over the five years since. Applied to that buy, the clause should be delivering about $60,000 a year in price relief right now. Ask any steel buyer whether their line item fell by $60,000, and enjoy the silence: the steel index held the 82nd percentile of its history while the ore clause pointed down. The failure runs in both directions. Had the same clause been signed ahead of ore's run to its June 2021 peak, it would have handed the mill an increase the correlation says its own costs did not require. Indexing a contract to a 2.0% explainer is not risk sharing; it is a coin flip with paperwork.

What to do with a 0.14

First, strike iron ore from the escalator language in your steel contracts. If a price must float, tie it to the output: the BLS iron and steel index itself, which is published monthly, is auditable by both sides, and tracks the thing you actually pay. Second, treat the next ore-citing surcharge letter as an opening position rather than a bill. Ask for the mill's realized cost stack, scrap, energy, and labor, and let the 0.14 do the arguing. Third, re-quote standing work against the current index instead of the one in last year's spreadsheet; an output sitting at the 82nd percentile of its history is a fact your own pricing should already reflect. Finally, run your last twelve months of invoices through the material price variance calculator against the published index. The gap between what the index moved and what your supplier moved is the number worth carrying into the next negotiation.

Put your invoice history and the live iron and steel index into the material price variance calculator to separate what the market moved from what your supplier moved. Audit the next surcharge

Published 2026-08-18.