How to count rocks
Sat May 14 2022
I spent a lot of this rainy Saturday on the slightly brain-numbing task of collecting and comparing different Harmonized System codes for various so-called “critical minerals.” HS codes are used to classify traded products: they’re maintained by a group called the World Customs Organization. The codes are divided into various “chapters” within which different sub-categories are listed. For example, chapter 26 is for “ores, slag, and ash” mineral products; section 2612 is for uranium or thorium ores and concentrates, and 261220 is for thorium ores and concentrates. Countries will use these high-level categories and also make up sub-categories at more granular levels for their specific needs: for instance, in the European Union 26.12.20.10 is designated for monazite ore, which is a source of thorium. Meanwhile, if you go to look up Brazilian trade statistics 26.12.20.10 isn’t a searchable category–monazite’s just lumped in with other thorium ores.
This may not be an especially significant distinction–monazite is the primary source of thorium ore and concentrates, so separating it out reflects a desire by the EU to keep close tabs on the mineral. It gets weirder with lithium, a that’s found in both mineral ore and in aquifers. “Lithium ore and concentrates” isn’t an available high-level HS code category; at best you can use 261790–the code which is likely noisy with other obscure ores. Lithium brine water was at one point classified by Customs and Border Patrol as “other mineral substances” (253090)–which maybe made sense in 1994 when lithium wasn’t a hot commodity, but makes less sense now.
But as with monazite, these classification issues might also not be too important. Most places that mine lithium (as brine or as rock) tend to refine in situ, so if you’re trying to find out which countries that mine lithium exported the most lithium using hydroxides and carbonates is a reasonable-enough proxy. So why am I doing this weird task of compiling HS codes beyond the fact my advisor asked me to?
Well, trade data about mineral commodities is one way that pepople make arguments about demand for particular commodities or argue that a material is uniquely “critical” or use it to create coarse proxies for measuring illegal mining activity. It’s hard to do any of those things precisely or with terrific confidence when you’ve got three different materials lumped into one category (as is the case with HS code 261590–” or vanadium ores and concentrates”). It’s not very useful for distinguishing which rare earth elements are the most “critical” when the rare earths are mostly lumped into one big category. The limitations of big trade datasets are generally known among the people who care about this sort of stuff, but not necessarily to the general public who hear a statistic about demand for tantalum and assume it’s correct. Also, there’s a difference between how things should be categorized and how they are categorized–apparently, it’s not unusual for monazite exports to end up misclassified as simple sands (250590–despite the fact the HS code description specifically says “Natural sands of all kinds, whether or not coloured (excl. silica sands, quartz sands, gold- and platinum-bearing sands, zircon, rutile and ilmenite sands, monazite sands, and tar or asphalt sands).” This seems to be the only time the word “monazite” appears in WCO-level HS codes?).
A dataset being not-so-secretly imprecise isn’t especially a huge surprise to anyone who’s worked with data in any form. And I worry about just how deeply pedantic all of this sounds when I try to explain it. For me the part of this stuff that’s interesting or important is more the larger project of classifying and categorizing the world into standardized commodities–the broader violence of creating a “normal” version of anything and how doing so facilitates higher-velocity movement of capital, while at the same time the mechanisms of capital remain fairly opaque to outsiders.