The Goldman Sachs Aluminum Conspiracy Was Pretty Silly
I am still digesting the giant Senate report on banks and commodities, but it's worth starting with the fun part: the Senate's version of the Goldman aluminum warehouse scam. Because, credit where it's due: The Senate's Permanent Subcommittee on Investigations really has advanced the ball on that. It's been over a year since the world first heard the mysterious story of Goldman taking aluminum out of warehouses to show it the sights of Detroit and then return it to the same warehouses, but its mysteries are only deepened by each new set of revelations. The Senate's are pretty good!
The basic story has always been that Goldman charges people above-market rents to store their aluminum in its warehouses and won't let them take it out. That sure sounds like a scam, but also a really dumb one. You can store aluminum anywhere; why store it in an overpriced terrible warehouse?
Amazingly, though, it seems to be true! Sort of. The Senate's report is fascinating but perilous reading, apparently produced with the benefit of subpoena power but without a firm understanding of commodity-market dynamics. I don't mean this as a criticism; the underlying commodity-market dynamics baffle me too. In fact, I get the strong sense that they baffled the people at Goldman who were actually doing these transactions, which is maybe the only way to explain how they happened. But let's press on nonetheless.
Here's the background.
You can store aluminum anywhere, in your garage if you want, it is the easiest thing in the world to store, it just lies there. But you can also store it in what is called "on-warrant storage" at a warehouse regulated by the London Metal Exchange. The metal in the LME warehouses can be used for delivery under LME futures contracts, so it is the foundation of the system for trading aluminum for financial purposes like hedging or speculation. "Off-warrant" aluminum, in garages or whatever, is better for making beer cans with. "There are currently approximately 4.4 million tons of LME warranted aluminum and there is estimated to be almost 8 million tons being stored off-warrant."
Goldman's Metro International Trade Services subsidiary owns a lot of the LME warehouses in the U.S. It also does off-warrant storage of aluminum. The rent at LME warehouses is higher than the rent at non-LME warehouses. The LME warehouses -- some of them, anyway, including the Metro ones in Detroit -- are also slow to give your metal back: If you want to take metal out of a warehouse, you send a notice to the warehouse (it's called "cancelling the warrant"), and then you wait in a queue for almost two years for your metal, paying the high rents all the while. The LME requires warehouses to load out at least a certain amount of metal each day, and the warehouses tend not to go much above this minimum, so the line is long.
So here's a simple question: Why would you store aluminum in an expensive warehouse that provides terrible customer service, when you could just store it in your garage? This is not a trick question, and it has an answer, and the answer is a good all-purpose answer to most questions, and it is: because someone pays you to do it. Metro would just call up people who had aluminum and say, "Hey do you want to store your aluminum at our warehouse?" and they would say, "No your warehouse is expensive and dumb," and Metro would say, "OK but we'll pay you to store it with us," and the people would say "How much?" and then they'd work out a deal.
(There's one other big reason to hold on-warrant aluminum, which is: It's easier to trade. If you want to get at your aluminum, Metro's LME warehouses are terrible. But if you want to get rid of your aluminum, LME storage is great: You can just deliver it into a futures contract, and never need to worry about putting it on a truck and shipping it to your buyer. The LME warehouses, as part of the system for trading abstract aluminum, do make trading much easier. )
Anyway, everyone involved is a rational actor, and you can sort of model up a rational deal. I mean wait no. You can model up kind of a crazy deal, but rational within the following parameters:
- Metro's LME warehouses charge rents that are higher than rents at non-LME warehouses.
- But Metro pays upfront cash incentives to make the warehouses attractive nonetheless.
- And they aren't term rentals: Customers can cancel their warrants at any time and withdraw aluminum; they just have to wait in the queue.
Those parameters are not obvious. You could imagine warehouses that give rent discounts rather than upfront incentives, or that require customers to store metal at the warehouse for a specific period and pay a break fee if they leave early.
But within Metro's system, you could create a rational deal, that is, a no-arbitrage deal where the cost of storing with Metro is the same as the cost of storing elsewhere. Here is roughly how it works:
- Metro charges an above-market rent and pays an upfront incentive.
- The upfront incentive is, roughly, the discounted amount of excessive rent that it expects to be able to charge over the period that the customer stores metal with Metro.
- That period is, in expectation, as short as possible: If you are a customer, and you're paid a one-time cash fee to store your metal in a warehouse that charges an above-market rent, then you should take the fee, cash the check and immediately take your metal out.
- Only "immediately" means "in like two years," because if you ask for your metal back it takes a long time to actually get out.
The point is that all of Metro's aluminum should be in the queue to leave its overpriced storage. But Metro knows that. Ignoring time value, Metro should offer customers an upfront incentive equal to
(Metro Rent - Non-LME Market Rent) x (Amount in Storage / Daily Amount Required to Load Out)
That is, if there were 300,000 tons of aluminum in Metro's warehouses, and it was required to load out 3,000 tons a day, then it should expect a new customer to pay 100 days of overpriced rent, and it should rebate the present value of that amount to the customer to get the customer to come in.
You can check this math, sort of, and the results are strange but revealing. Page 187 of the Senate report says that Metro's Detroit LME aluminum stock was about 1.5 million metric tons. As of that time, the queue to remove metal from those warehouses was more than 600 days (page 193). The LME rules required Goldman to load out 3,000 metric tons of aluminum per day at that time (page 191). If you multiply those last two numbers you get 1.8 million tons of aluminum in the queue. But there were only 1.5 million tons in the warehouses.
That is: More than all of the aluminum in Goldman's LME warehouses was trying to get out! That can't be right -- there's some sort of glitch somewhere in those numbers -- but it seems close to right. Roughly speaking, everyone who stored aluminum with Metro wanted to take it out. [Update: Goldman informs me that the glitch in the numbers is that the LME load-out requirement is not always 3,000 tons per day; the requirements are graduated. So for instance in October 2013, the queue was 837,000 tons of aluminum, or 445 business days (623 calendar days) at around 1,900 tons a day. That's versus around 1.5 million tons of aluminum in the warehouse. So the real number is: more than half of the aluminum in the warehouse was trying to get out. The percentage seems to have gotten even higher in early 2014. The rest was apparently being used for trading, etc.]
People were clamoring to get aluminum out, but they were putting even more in. Within this goofy system, that is perfectly rational: Goldman's warehouses were bad places to keep aluminum, but pretty good places to put aluminum into. Because you got paid to put aluminum in, and then you overpaid to keep it there.
OK now the scam! Here it is, from pages 194-195 of the Senate report:
Metro’s merry-go-round deals took place in 2010, 2012, and 2013. According to a Metro executive, the deals began in the summer of 2010, just a few months after Goldman acquired Metro, when Metro became concerned that owners of aluminum in its warehouses were removing the metal from its warehouses and storing it elsewhere, leading to a loss of revenue. In an effort to curb that loss, Metro executives and the Metro Board of Directors, composed exclusively of Goldman employees, made a strategic decision to -- for the first time -- “market” Metro incentives to metal owners that already had metal stored in Metro’s warehouses. ...
In each deal, Metro provided financial incentives to the owner of the aluminum stored in its warehouses to: (1) wait in the queue; (2) upon reaching the head of the queue, load out its metal from a Metro warehouse; (3) deliver the metal to another nearby Metro warehouse; and (4) warrant the metal while in the second Metro warehouse. Each deal led to aluminum being loaded out of one Metro warehouse in Detroit and loaded right back into another, a practice that one Metro forklift operator later told the New York Times amounted to a “merry-go-round of metal.”
Metro was just cutting out the middleman. The rational thing for all of its storage customers to do was to take their metal out, store it in their garage for a little while and then put it back into the Metro system when Metro came calling with big upfront incentives. Metro just said, "Well, OK, if you're going to do that, why don't we arrange it all in advance, so you get the incentives before you even take the metal out?" And that's what happened.
So the scam was real! But was it a scam? I mean. It's really dumb, right? Surely you could imagine a better system than overpaying customers to put aluminum into a warehouse where they overpay to keep it there, and then watching them immediately try to take it back out again and wait two years to do so. Or don't imagine it, it happens: "Of the LME warranted aluminum, more than half is held at 'non-queue' locations" that have solved this entirely self-created problem. Like: When Deutsche Bank or Red Kite or whoever came to Metro and asked to take their aluminum out, Metro could have said, "No no we'll give you a discount on rent if you agree to leave it here for a while." Instead it said, "OK fine, take it out, but we'll pay you to bring it right back again." That is pretty clearly worse.
But worse for whom? This process was dumb, but who suffered from it? Metro's customers don't seem to be complaining, presumably because they all knew the (dumb) rules by which they were playing. It seems to be the case that basically everyone was always in the process of removing their metal from Goldman's warehouses, extremely slowly, and then loading it back in when the incentives were right. This dumb system cleared a market, and the participants in the market seem to have just gotten used to it.
The people who are complaining are the people who think that the dumb system pushed up the price of aluminum: Because customers couldn't get aluminum out of Metro's warehouses, there was a shortage of aluminum and the price went up. This story is exactly false: There was a glut of aluminum, and the price went down. Here's a price chart from page 173 of the Senate report:
The red line, on the other hand, is the premium over the LME price that you pay to get actual aluminum actually delivered to you in the Midwest -- basically, the price you pay to get aluminum today instead of when it comes out of a Metro LME warehouse in two years. That premium was going up, because LME aluminum became less valuable relative to actual aluminum, because it took so much longer to get.
If you just look at the premium -- the red line -- then you might think that Goldman's schemes were driving up the price of aluminum, and the Senate seems to believe that a bit, though I don't especially. On the other hand, this is more compelling (page 180):
The investigation also found that the price impacts of the queue had created problems for aluminum users like beverage can producers and automobile manufacturers who actually use aluminum, because the increasing difference between the all-in price and the LME futures price made hedging price risk through the LME market increasingly ineffective. A number of commercial users told the Subcommittee that the lack of effective hedges damages planning and impacts revenues.
When we last talked about the great aluminum conspiracy, I mentioned this problem: that the price of financial, abstract aluminum -- the stuff stored in the LME warehouses and used in futures trading -- was becoming disconnected from the price of real, available aluminum that could be made into beer cans. But I said that that was not a problem for beer-can prices, because beer cans are made with real aluminum, not abstract LME aluminum. But this is a fair point! Beer-can makers don't just make beer cans. They also make financial decisions, including decisions about hedging their aluminum costs. And when abstract aluminum and real aluminum diverge, then the costs and risks of that hedging go up. And that does cost the beer-can makers.
I don't know that Metro's "merry-go-round" deals actually caused that divergence. The Senate implies that they did, but I'm not sure. It seems reasonable to think that the enormous queues to take aluminum out of Metro's Detroit LME warehouses helped cause the widening divergence between the LME price and the available-metal price: Metal that you can use now is obviously more valuable than metal that you can't get at for two years. I'm less convinced that the merry-go-round trades increased the queues; on my reading, it seems like the queues would have been long no matter what, because customers' incentives were to unload as much aluminum as possible, and Goldman's incentives were to unload it as slowly as possible.
But that's still bad! That has nothing to do with scams or schemes or secret back-room deals, and it was all disclosed and agreed and within the rules. It's just that the rules were inefficient and the incentives were bad. It's dumb, and it's dumb in a way that seems to have made markets less efficient and hedging harder and, sure, maybe, beer more expensive.
(Updates paragraph 14 with Goldman Sachs's comments on load-out rates.)
DISCLOSURE: I used to work at Goldman Sachs, though not in anything commodities-ish. I still own a little restricted stock, and earlier today I was down tens of dollars, perhaps due in part to the Senate's work. It got better though.
We've discussed this before in some detail. Much of my knowledge, alas, draws on a Goldman Sachs Commodities Research report ("The economic role of a warehouse exchange," Oct. 31, 2013), so I guess there might be some embedded biases there.
That's from a Goldman commodities trader's testimony today, so salt it however you'd like.
The May 2014 peak was, as the Senate report puts it (page 193), "a stunning 674 days."
Because they can charge the high rents, while they wait. There is also an argument that the LME stocks should be "sticky," since it serves as sort of a last-resort reserve of aluminum, and it's useful to have stable stocks.
The Senate's version (page 188):
Metro’s increasing budget allocation for aluminum freight incentives supports that analysis. In early 2010, just after Goldman acquired the company, Metro paid nearly $37 million in freight incentives to attract aluminum to its warehouses. That figure doubled in one year to nearly $79 million in 2011, grew to nearly $103 million in 2012, and reached nearly $129 million in 2013, an increase of nearly 350% over four years.
Relatedly, a reason to store aluminum at an LME warehouse is that you got delivery on a futures contract, and so now you're stuck with that aluminum until you take it out.
HUGE OVERSIMPLIFICATION. For one thing, there is a load-out charge to take your metal out, so you have to amortize that too, and it cuts the other way. For another thing, like I said, there are non-incentives-related reasons to want to keep metal in an LME warehouse. Like, you might want to sell it. Also it looks like Metro's incentives were not just up-front cash payments, there were discounts and other things as well. I think the story I'm telling in the text is the right story in broad outlines, but only with like 80 percent confidence, and the details blur it a bit.
Between 2003 and 2011, the LME’s minimum load-out rate was 1,500 metric tons per day for the largest LME warehouses, such as Metro’s Detroit warehouses. In April 2012, the LME increased that number to a rate ranging from 1,500 to 3,000 metric tons a day, depending upon a warehouse’s closing stock level. In November 2013, the LME adopted a rule that would have linked a warehouse’s load-in rate to its load-out rate as of April 2014, but the rule was subjected to a court challenge. Metro nevertheless began voluntarily complying with the new rule in April.
That seems to be per warehouse, but (page 195) "Metro used a single exit queue for all of its Detroit warehouses combined."
Page 187 of the Senate report:
Whereas in January 2008, less than 400,000 metric tons of LME warranted aluminum were in storage in the entire United States, by the end of February 2010, Metro’s Detroit warehouses alone were storing about 915,000 metric tons. Over the next two years, Metro’s Detroit aluminum stocks continued to grow, reaching about 1 million metric tons in January 2011, and about 1.4 million metric tons by February 2012. A year later in 2013, they remained at nearly 1.4 million metric tons and, by February 2014, Metro’s Detroit aluminum stocks stayed steady about 1.5 million metric tons, nearly all of which was on LME warrant.
Goldman's response has a sort of nah-nah-nah-we-can't-hear-you flavor; here is the statement of the chief executive officer of Metro:
I appreciate the opportunity to describe the off-warrant movement of metal, which has been characterized incorrectly. Metro offered customers that were removing or considering to remove metal from its Detroit warehouse the opportunity to store the metal off warrant in a different Metro warehouse. Such customers had various other options, including storing their metal with competing companies, many of which have warehouses near to Metro’s. Metro offered these off-warrant transactions to compete for the continued storage of this metal, but it is always up to the owner, not Metro, to decide what to do with its metal.
The metal at issue in these relatively few transactions was loaded out by Metro at the owner’s instructions onto a truck, issued a bill of lading, and moved to another location at the owner’s direction. Once the owner made this choice, the LME rules required that Metro follow the owner’s instructions regarding the disposition of its metal, including treating this metal as loaded out and reducing its LME inventory stocks accordingly. The fact that the owner moves the metal between two Metro warehouses in the Detroit area is no different under the LME rules than if the owner moves the metal to an equally close non-Metro warehouse.
But the whole point, voluminously documented in the Senate report, is that the metal was not moved to off-warrant storage, except briefly, and that it didn't reduce Metro's LME inventory. Instead, the metal was quickly re-warranted at the other Metro warehouse, and everyone knew all along that that was the plan. From the Senate report (page 199):
The first three deals with Red Kite took place from January through March of 2012. In those transactions, Metro offered financial incentives for Red Kite to cancel warrants on a combined total of 250,000 metric tons of aluminum, wait in line, load out the metal from Metro warehouses, load it back into other Metro warehouses, and re-warrant the metal. The incentives offered by Metro included: (1) paying a “day one” cash incentive to the metal owner when the metal warrants were cancelled, (2) offering a period of free rent, and (3) paying another cash incentive for re-warranting. As in the Deutsche Bank deal, each transaction required Red Kite to pay a substantial cash penalty to Metro if Red Kite did anything other than re-load the metal into a Metro warehouse and re-warrant it.
As for that Deutsche Bank deal (page 198):
Of the original 100,000 metric tons of aluminum subject to the deal, approximately 70,000 metric tons left one Metro warehouse for another Metro warehouse in Detroit, and were then re-warranted. The remaining 30,000 metric tons were placed back on warrant before they were actually loaded out. Thus, in the end, all 100,000 metric tons were back on warrant at Metro at the end of the deal.
Like every other market, the price of aluminum is established through supply and demand fundamentals. And, those trends have been unmistakable. There has been a consistent surplus of aluminum since 2008, resulting in a large volume that has been placed in storage. Each year, approximately 49-50 million tons of aluminum are produced. Since 2008, production has exceeded consumption by one to two million tons a year, resulting in an increasing surplus that has gone into storage. That’s why there has never been a shortage of aluminum.
Would it have gone down faster if they hadn't? Probably not? Like, you have an economy. There is some amount of demand for aluminum. You can store your aluminum, or you can sell it. If you want to sell, you sell. If you think it'll be worth more later, you store it. All of those are decisions driven by fundamentals. It's true that lower prices now, and an expectation of higher prices later, create an incentive to store (or not mine) aluminum, instead of selling it right away and driving the price down further. But that's not a story of merry-go-rounds or incentive fees. That's a story of supply and demand and expectations.
If you have all of the aluminum, or even most of it, then you can hoard aluminum and create scarcity and manipulate the price and all that. But Goldman, or even Goldman plus its customers, didn't. The Metro warehouses had 1.5 million metric tons of aluminum in 2014, a few weeks' worth of aluminum production.
Here is a bad passage (page 180):
In contrast, the LME and Goldman contend that longer queues have not affected the all-in price for aluminum. Although both the LME and Goldman concede that the queue has affected premium prices and the relative proportions of the all-in price attributable to the premium price versus the LME price, they assert that the effect of the longer queue has been to drive the LME portion down and the premium portion up, leaving the all-in price substantially unchanged. That analysis is a minority view, according to briefings provided to the Subcommittee by numerous aluminum market participants and experts. Alcoa, the largest U.S. aluminum producer, told the Subcommittee, for example, that the LME and premium prices are not inversely related, but move independently of one another. In a recent filing with the SEC, Alcoa wrote that the LME price and the aluminum premium each “has its own drivers of variability.” Mr. Vazquez, the aluminum analyst, agreed with that view, indicating to the Subcommittee that “there has been no empirical study or evidence or modeling that suggests changes in LME prices and the Midwest Premium are inversely related,” as the LME and Goldman have suggested. In fact, the LME and Midwest Premium prices can and often have moved in the same direction.
Obviously they're not "inversely related"! The aluminum price moves. If the premium and the LME price were precisely inverse, then the all-in price would always be the same. But it isn't! In fact, over this period, the all-in price went down.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To contact the author on this story:
Matthew S Levine at firstname.lastname@example.org
To contact the editor on this story:
Zara Kessler at email@example.com