So much aluminum. Photographer: Andrey Rudakov/Bloomberg
So much aluminum. Photographer: Andrey Rudakov/Bloomberg

Let me tell you about aluminum. Aluminum is a metal that is used to make beer cans, and also for some other things. You might want to buy aluminum for either of two reasons:

  1. to take that aluminum and turn it into beer cans, or
  2. for "financial purposes."

One big financial purpose is to hedge your future plans to make beer cans: You think aluminum looks cheap today, so you buy some aluminum to lock in your costs. But there are other possible financial purposes. Maybe you are hedging some risk in some other commodity that is correlated with aluminum. Maybe you are just speculating on the price of aluminum. Whatever. Financial purposes are broad; the point of this category is that you want exposure to the price of aluminum, as opposed to an actual pile of aluminum to be fed into your beer-can-making machine.

Similarly, you might sell aluminum because it just came out of your aluminum mine1 and you want to get rid of it. Or you might sell it for financial purposes, hedging your future aluminum mining or whatever.

So there are sort of two markets for aluminum.2 One is, like, A digs aluminum out of the ground and sells it to B to make beer cans. The other is, like, C and D speculate with each other on the price of aluminum. These two markets serve different, though obviously related, purposes. The beer-cans market is about putting a big hunk of metal on a truck and delivering it to the gate of your factory, where you will turn it into beer cans. It is concrete and specific and local: You need a specific amount of aluminum, in a specific shape, at a specific place and specific time.

The financial market is about figuring out the price of aluminum, balancing worldwide supply and demand, moving aluminum to its highest and best uses by trading away arbitrage opportunities. It abstracts away from shapes and places and times and just trades aluminum generally. So everyone agrees that there's a thing called The Aluminum Price, and it's the price of a standardized amount of a standardized grade of aluminum delivered to an abstract location.

How do you deliver actual metal to an abstract location? Well, you put a bunch of aluminum in special warehouses around the world, and everyone agrees that the insides of those warehouses constitute Abstract Aluminum Space. So if you buy aluminum for hedging or other financial purposes, then what you have is not a lump of aluminum on a railroad siding by your factory, but rather a claim on some aluminum that lives in Abstract Aluminum Space. And when you later want to close that position and take the gain or loss on your financial investment in aluminum, you sell the claim on to someone else. You never bother moving actual lumps of metal around; you just trade the abstract claims.3

Of course, you cannot make actual beer cans with abstract aluminum, and for the abstract market to be useful in setting the price of aluminum, there needs to be some interchangeability with the actual-lumps-of-metal market. There is, but it's annoying. If you've bought abstract aluminum, you can ask to get the actual underlying aluminum. The seller will give you what is called a "warrant," which entitles you to take out a particular amount of aluminum out of the network of warehouses that constitute Abstract Aluminum Space. But he doesn't have to give you a very convenient warrant: If your factory is in Detroit, he can give you a warrant for metal in Barcelona.

And once you have the warrant, it's not very convenient to get the aluminum out. The Abstract Aluminum Space warehouses exist not to facilitate industrial processes, but to facilitate abstraction. What that means is, basically, it takes a really long time to get aluminum out of them. A forklift driver will work with less urgency if he is an agent of abstraction than if he is an agent of industry.

But this shouldn't bother you. You bought the abstract aluminum for financial purposes. Let's say you bought it in January to hedge your aluminum-price exposure. In September, when you're actually going to make beer cans, you don't take your aluminum out of the abstract system and have it delivered to your factory in Detroit. Instead, you sell your abstract aluminum to lock in the profit or loss on your hedging trade, and you buy actual aluminum for delivery to your factory in Detroit. The abstract aluminum remains in the abstract system, and you've just used it for its price. You buy your actual aluminum from an aluminum smelter, or from a warehouse that isn't part of the abstract system.

The actual aluminum (which can be made into beer cans) is more valuable to you than the abstract aluminum (which can't), so you pay more for it. If you're in Detroit, this is called the "Midwest Premium." You can think of that premium as "the price of converting from abstract to actual aluminum," if you want, but you don't have to. You can just think of it as: There's a market for people to buy and sell hunks of aluminum in Detroit, and there's a different market for people to buy and sell abstract aluminum claims in Abstract Aluminum Space, and those markets have different prices. But the point of the abstract market is to help set and hedge the price in the local market (and lots of other local markets elsewhere), so if they get too far out of whack -- if the Midwest Premium gets too high, or fluctuates too much -- then the abstract market isn't doing its job well. The more closely the local price tracks the abstract price, the better the abstract market is doing.

That was the story of aluminum. I hope you enjoyed it.

There is, however, another story of aluminum, or maybe it's the same story in different words. About a year ago, the New York Times published a famously puzzling article about Goldman Sachs's network of aluminum warehouses.4 The point of the story was that Goldman -- in its Metro International subsidiary, which it's now selling -- did an evil trick where it would load aluminum onto a truck, drive the truck around Detroit, and then deliver it back to the warehouse. This apparently made a lot of money for Goldman, though the mechanism was a bit mysterious.

The idea seemed to be that Goldman's warehouses were very expensive places to store aluminum, and that by never letting the aluminum leave the warehouse except for short sightseeing trips around Detroit, Goldman could keep the aluminum in the warehouses, charge a lot of rent, and create an artificial scarcity of aluminum. This was bad for beer companies, who just wanted their aluminum so they could make cans to put your beer in. The Times:

Hundreds of millions of times a day, thirsty Americans open a can of soda, beer or juice. And every time they do it, they pay a fraction of a penny more because of a shrewd maneuver by Goldman Sachs and other financial players that ultimately costs consumers billions of dollars.

Some aluminum consumers read the article and duly sued Goldman. On Friday, a federal judge in New York dismissed the lawsuit.

That seems about right. There's an obvious problem with the theory that Goldman was scamming people by never letting anyone see their aluminum. No one needs to pay Goldman to store aluminum in expensive slow-moving warehouses. I said once that aluminum "is a metal, nonperishable, not especially fragile and unlikely to run away. You could store it in your garage." So the idea that Goldman would run a successful aluminum-storing business that distinguished itself with (1) high prices and (2) terrible service never made a ton of sense. Just use a different warehouse!

The mystery clears up a bit when you remember Abstract Aluminum Space. Goldman's controversial warehouses were part of Abstract Aluminum Space. More technically, they were warehouses regulated by the London Metal Exchange for "on-warrant storage," that is, storage of aluminum that can be traded using LME contracts, which are the main tool for trading aluminum for financial purposes. If you didn't care about trading abstract aluminum using LME contracts -- if you wanted to store some aluminum now to make some beer cans later -- then there was no real reason to use those warehouses. From Goldman's legal filings:

Because aluminum will not degrade if stored outside -- and, unlike other metals, is not subject to serious theft risk given its value relative to its weight -- off-warrant aluminum can be stored essentially anywhere at relatively little cost. ... Because off-warrant storage is not subject to the same regulations and does not provide the same liquidity as LME warrants, rent for off-warrant storage is substantially less (often ten times cheaper or more) than LME rates for on-warrant storage.

And so in fact more aluminum (7 to 10 million tonnes) is stored at non-LME warehouses than at LME ones (5.5 million tonnes). Really, the natural question is: Why would anyone store aluminum at Goldman's inefficient and expensive on-warrant warehouses?5

One answer is, because they like trading abstract aluminum. But another answer is, because they're paid to. Remember, Abstract Aluminum Space warehouses exist to facilitate abstraction. The LME, and abstract-aluminum traders generally, want there to be aluminum in Abstract Aluminum Space. Here's how they do that:6

[I]n order to compete with off-exchange storage, the LME warehouse offers inducements to attract metal at the beginning of the cycle, but also charges exit fees and has other barriers to withdraw to keep the metal until the end of the cycle, while also creating a secure credit and quality environment.

But you could read that in a nefarious way. From the judge's opinion dismissing the case:

Goldman/Metro offered to pay as much as $250 or more per ton in such up-front “incentive payments.” According to plaintiffs, Goldman’s incentive payments created a positive feedback loop in which the more aluminum it could divert into its warehouses and the more inefficient it could be in loading-out, the more it earned and therefore the more it could afford to pay in incentive payments.

On this reading, the scam was:

  1. We'll pay you to store aluminum with us.
  2. You'll pay us rent to store that aluminum.
  3. We'll take our sweet time in giving it back to you.

Which means that Goldman was not scamming its customers. Rather, Goldman and its customers were working together to scam unrelated aluminum consumers. The people who sued Goldman weren't customers who couldn't get their aluminum out. They had no relationship with Goldman. Rather, they are manufacturers who bought "primary aluminum ingot" and "primary aluminum billet" from smelters and other aluminum sellers. That is, they traded entirely in the local, actual aluminum market, not the abstract one. Their claim isn't that Goldman overcharged them, or didn't give them their aluminum when they asked for it. Their claim is that Goldman and its customers pushed up the price of aluminum for everyone who didn't deal with Goldman.

Now this is obviously true. Goldman's customers bought aluminum and stored it. If they hadn't bought it, or had sold it instead of storing it, then the price of aluminum would be lower. Buying makes prices go higher. Selling makes them go lower.

But generally, you're allowed to buy stuff, even if buying stuff makes the price go up. You're not allowed to conspire to manipulate prices, but just buying stuff isn't a conspiracy to manipulate prices. As the judge put it in dismissing the case:

[T]he allegations tell a story consistent with market-driven behavior by traders and warehouses rather than unlawful conspiracy. It does not make economic sense -- that is, it is not economically plausible without more -- for defendants to have conspired to achieve an end that the law of supply and demand, combined with the passage of time, would have itself achieved in the absence of a conspiracy.

There's no conspiracy. People bought aluminum because they wanted to hold aluminum -- in part because the price of aluminum was really low.7 Then they stored the aluminum. That pushed the price of aluminum up, of course. But there's nothing illegal about that. That's just a market working.

This is all incredibly simple but it's confused by one issue, which is the difference between the abstract aluminum price and the actual price for delivery in Detroit or wherever. The lawsuit doesn't say that Goldman's customers pushed up the "price of aluminum." Rather it says that Goldman and its customers pushed up the "Midwest Premium," the difference between the actual price of aluminum in Detroit and the abstract LME price.8 And while aluminum prices remain relatively low, the Midwest Premium was unusually high during the class period.

If you're conspiracy-minded, you will assume that the LME price -- the abstract financial price -- is the "price of aluminum," and that the Midwest Premium represents the cost of converting abstract aluminum into actual aluminum in Detroit. When that premium goes up, despite a glut of aluminum in Detroit, you will naturally find that suspicious. And the fact that getting aluminum out of Goldman's warehouses takes forever will add to your suspicions.

But you don't have to think of it that way. You could say: The price of aluminum in Detroit is the price of aluminum in Detroit, and it's set by supply and demand for aluminum in Detroit. The LME price is the price of aluminum in Abstract Aluminum Space, and it's set by supply and demand for aluminum in Abstract Aluminum Space. If you subtract those two prices, you'll get a number, and that number is called the "Midwest Premium," but it's just a number. If it goes up, that just means that aluminum in Detroit is becoming relatively more valuable than aluminum in Abstract Aluminum Space.

That latter reading happens to be right. The Midwest Premium is not a separate cost but is rather the arithmetic result of subtracting real prices observed on the LME from real prices observed in the Midwest market.9 The fact that the premium went up doesn't mean that the price of aluminum in the Midwest was too high. It means that the price of aluminum in Abstract Aluminum Space was too low: that it was becoming a less effective mirror of the real cost of aluminum.

That's a problem for the abstract aluminum market, and one that it's collectively trying to fix.10 But it's not a problem for beer-can prices. Goldman and its customers can't make aluminum prices in the Midwest go up just by charging each other fees and moving trucks around. Your beer is safe.

1 Actually your bauxite mine, or your smelter. But we are being nontechnical here, like we were with the beer cans.

2 This version of the story, which is abstract and stylized, draws on a couple of key sources. One is a Goldman Sachs Commodities Research report, titled "The economic role of a warehouse exchange," from Oct. 31, 2013, sadly not available online though it's excerpted here. Or call your Goldman salesperson and tell them to make it public. Goldman obviously has something other than a purely intellectual interest in this story, so you could take this with a grain of salt if you want.

(You can take me with some salt too, if you want! In this post I will defend Goldman Sachs so I'll remind you: I used to work there, though not in the aluminum business, and I still own a bit of restricted stock.)

The other sources are the documents in the aluminum warehousing antitrust case, particularly:

3 This sort of abstraction is of course common in financial markets. When you buy and sell stock, you don't exchange stock certificates; you exchange claims on stock certificates that are held centrally at DTCC.

4 I mean, I sure found it puzzling, as I've mentioned before (repeatedly). Matt Yglesias found it very puzzling too:

And yet having read the piece twice, I don't understand how the scam works. The basic idea seems to be that by shuffling aluminum around rather than delivering it promptly, Goldman can charge more rent and boost its profits.

But this (a) doesn't appear to have anything in particular to do with Goldman Sachs' well-known investment banking activities and (b) sounds like far too ridiculous a scam to actually work. It amounts to saying that Goldman's amazing business strategy is to deliberately provide terrible customer service (shipping delays) and high prices (charging extra rent to cover the delays). If it were that easy to make billions of dollars, we'd all be doing it. The story is entirely missing a clear explanation of why this doesn't just lead someone else to open aluminum warehouses and undercut them.

Kevin Drum also found it confusing.

5 Incidentally, Goldman also had some off-warrant warehouses, which led to trouble:

In some instances, Metro has attempted to compete with these non-LME warehouses by offering warrant holders the opportunity, after they have cancelled their warrants and waited in whatever queue existed at the time, to move their aluminum to a different Metro warehouse in Detroit for off-warrant storage at a reduced rent with the option of re-warranting the metal should they choose to do so. This competition has resulted in the so-called “shuttling of metal from one warehouse to another” to which the complaints refer.

6 From that Goldman research report I mentioned in footnote 2.

7 Here's how the motion to dismiss starts:

The financial crisis of 2008 and the global recession that followed led to a sharp drop in industrial demand for aluminum. Because aluminum producers around the world did not correspondingly curtail supply by shutting down smelters -- which are very expensive to restart -- a huge global surplus of aluminum developed that continues until this day. As plaintiffs themselves acknowledge, global production of aluminum has exceeded consumption each and every year since 2008, resulting in a massive stockpile of surplus aluminum. Sluggish demand and excess supply also have resulted in plummeting prices. More than five years after the height of the financial crisis, the price of aluminum still has not recovered; it remains nearly 30% below its peak in 2006.

On the other hand, aluminum is in steep contango, as the aluminum industry cuts output and expectations for future prices are higher than current prices. So one clever financial strategy would be to buy aluminum now and put it in your garage until prices are higher. An equally clever strategy would be to buy aluminum now and put it in a Goldman on-warrant warehouse until prices are higher, if Goldman gives you enough incentives to make on-warrant storage economically equivalent to storing the aluminum in your garage.

Izzy Kaminska at FT Alphaville is the leading writer on the aluminum warehousing trade; here is her piece from last July. The technical explanation is heavy on contango and financing, but it's unnecessary in the text here. The point is: If aluminum prices are low now, and you think they will go higher, you should buy aluminum now. But then you have to put it somewhere.

8 From the complaint:

In the Midwest and other parts of the United States, Plaintiffs and other first level purchasers of aluminum from smelters (or from Defendants or co-conspirator parties to Defendants’ unlawful agreements in unreasonable restraint of trade), have paid a common, standardized benchmark price consisting of the LME price plus the Midwest premium or Platts MW premium (or a premium described by similar names. The sum of these two payments and any additional premium has been called the “Midwest Transaction Price.” ...
However, due to Defendants’ abuses of their monopoly powers and anticompetitive agreements to inflate prices and supplies, the Midwest premium price has followed a historically anomalous pattern during the Class Period. In fact, it has behaved in exactly the opposite manner that supply, demand and transportation costs indicated. First, the Midwest Premium did not decrease. Second, on the contrary, it increased to all-time record levels.

And here is David Kocieniewski, the Times reporter who wrote the aluminum-shuffling story:

There are two components to the price of aluminum: the underlying cost of the metal plus a premium added on for storage and transportation costs. The warehouse delays at Goldman's facilities in Detroit inflated that premium, which has more than doubled since 2010. And the premium is added on to every ton of aluminum sold, whether the underlying price rises or falls.

Even companies that buy their metal elsewhere -- like Coca-Cola and MillerCoors -- have to pay the higher premium and pass that cost on to consumers.

9 From the motion to dismiss:

The difference between the “all-in” price and the LME spot price is referred to as the “premium.” A company called Platts calculates something called the Midwest Transaction Price “based upon what is described as ‘a daily survey of spot buyers and sellers, using a representative sample of producers, traders and different types of end users.’” According to plaintiffs, “[u]sing this survey information, Platts calculates a single price for the metal” -- an “all-in” price. “The difference between this single, all-in spot price and the LME price is the premium, which Platts then publishes.” Plaintiffs allege that through this subtraction exercise (“all-in” price minus LME spot price), Platts “derive[s] premiums for different markets. The U.S. premium is referred to as the Platts MW or Midwest Premium.”

Over the last several years, the LME spot price for aluminum has declined relative to the “all-in” price causing the two prices to diverge. As plaintiffs observe, “[s]tarting in February 2011, . . . the LME Cash Price began a decline from $2,600 to a recent low of $1,730.” These prices are down from a peak of “about $3,000 per tonne” before 2008. Because the Midwest Premium is the difference between the all-in price and the LME spot price, this sharp decline of the LME spot price relative to the all-in price has caused the Midwest Premium to “move[] in the opposite direction.”

Now I will say this for the plaintiffs: They allege that they had contracts with prices for aluminum calculated based on the LME price plus the Platts Midwest Premium. So for them, the Midwest Premium was a separate thing, and they were sad when it went up. But it's not a specific cost of getting metal out of warehouses; it's just an arithmetic residue.

10 Basically by trying to make its warehouses do a better job of delivering actual metal to people if they want it. There's a tradeoff: Making abstract aluminum more interchangeable with real local useful aluminum will reduce the premium and make abstract aluminum a better mirror of real aluminum, but reducing the amount of aluminum in Abstract Aluminum Space may reduce the liquidity and robustness of that market. From that Goldman research report:

Like other exchanges with warehousing systems such as corn and coffee, the economic role of the LME is to provide the benefits of liquidity, transparency and a “market of last resort”, with the primary cost being basis risk, the spread between the physical and exchange price, which in LME speak is “physical premium.” ...

Policies aimed at reducing the queue and lowering basis risk (premium) will impact the economic role of the exchange market, but have no impact on the economic role of the physical market. Such policy shifts may reduce basis risk with the cost of reducing transparency, liquidity and impacting the ability of the LME to create a market of “last resort”.

To contact the writer of this article: Matt Levine at

To contact the editor responsible for this article: Tobin Harshaw at