There are a few trades that are good enough to make it into a book, like taking the other side of Amaranth's natural gas trade in 2006 or betting on persistently low rates after the crisis.¹ Going long vol ahead of a panic even has an airport novel. But the trade that spawned multiple books, and a movie, was buying credit default swaps on subprime mortgage-backed collateralized debt obligations. (Having established a sizable short position in polysyllabic latinate jargon, I'll call these CDSs on CDOs from here on.)

This was, in dollar terms, exactly what the best book about it says on the tin: The Greatest Trade Ever. But it's really several trades in one. You have the directional bet that housing is overvalued and is going to go down. There's the asset class expressing the bet—it's hard to short individual houses, homebuilders are cyclicals with low peak-of-cycle P/Es so you don't get much juice there, so you'd want to bet against mortgages. And conveniently, credit default swaps had grown from a kind of financially-engineered diplomacy—banks sometimes regretted specific loans they'd made, but wanted to preserve a counterparty relationship, so they'd buy insurance on the loan from another bank—basically the gap between what the loan's spread and the CDS cost was an annual subscription to remaining buddies with management. And, at first separately, bankers were packaging together piles of loans into CDOs.

The way to think about a CDO is that it's a bank's balance sheet, just without a bank attached. What you do is collect a large number of loans—similar enough that your product fits in some coherent category, but diversified enough that, importantly, they don't all default at once. And then you sell debt against it. Suppose your CDO is composed of loans that would, individually, be rated around BBB, so an expected one-year default rate of around 0.5%. And suppose their defaults are not especially correlated. If you have $100m of them, you might end up selling $70m worth of AAA-rated paper, all backed by whichever 70% of them turn out to default last. You'd sell some smaller slices at other ratings, but you run into a bit of a problem there: at some point, the slice you have left is basically guaranteed to default, because you're asking "what's the performance of the worst loan in a portfolio where they're all pretty mediocre?" and the answer is that it returns roughly zero. So what you do instead is create an equity tranche, which gets whatever is left over if every other tranche is paid off. Or, equivalently, the tranche that takes the first loss any time there's a default.² That slice is pretty small; in this CDO prospectus from 2007, that piece is 4% of the $1.5bn structure, while 68% is AAA-rated or "Super-Senior," designed to be even safer than AAA.

If you put together CDOs and CDSs, you get a fun little opportunity: the interest paid by a given tranche, or the cost of buying insurance against it, implicitly sketch out the market's claim of how well the assets are correlated. If the correlation is exactly zero, that CDO is actually making the world a wealthier place: it's taking a bunch of different products that would have uncertain returns, and instead delivering investors their blended return, smoothing out the variance. And those investors get to choose exactly where on the risk/reward curve they want to sit. If correlations are 1, every tranche performs identically; either everything pays off, and a residual goes to the equity tranche, or everyone, from the equity to the super-senior gets wiped out, so they all deserve the same yield.

Both of these are extremes, but they illustrate something important: if correlations between defaults within a given CDO are higher than the market expects, the most senior tranche is less safe than it looks—but the equity slice is actually more safe than it looks. Correlated defaults can mean unexpectedly high default rates, but also unexpectedly low ones.

If you looked at historical data about mortgages, you got a certain view of how likely it was that a default in Florida predicted one in Maine. But if you considered how fast the subprime mortgage business was scaling, how fast some metro areas were growing, how dependent these economies were on homebuilding, etc., you could easily come to the view that local housing markets were now more correlated, and more driven by the availability of credit. That pushes correlations up, both for upside and downside scenarios: the good times stay good, and the most levered and thus least-creditworthy borrowers are the biggest beneficiaries when more credit availability makes their homes worth more and gives them the opportunity to refinance. And, of course, if it gets hard to get a mortgage, suddenly home prices will collapse just about everywhere, at the same time.

This setup means that you can structure the whole trade so that you're earning positive carry: your equity slices are paying you more than you're spending on insurance against the super-safe ones. This solves for a problem that some of the subprime short trade narratives underrate: figuring out the trade involved predicting the right trend and finding a good instrument to express that view, but maximizing the size of the trade at the optimal time was really an investor relations problem. John Paulson solved this one by creating a strategy-specific fund, so his investors were true believers. Michael Burry solved this with what his investors probably thought was a classic case of macro tourism—most of the time, "good value investor" plus "macro view expressed through complex derivatives" equals "world's best counterparty."³

If you're making a bet like this, you're making a meta-bet on timing. Plenty of people talked about housing being a bubble; by 2005, you could watch Flip This House or Flip That House. Being early means bleeding money for a while, and it's very hard for investors to underwrite a negative-carry strategy like this. So the dilemma for fund managers was that if they made the bet too late, they'd miss out, but if they made it too early, they might get preemptively cashed out. A positive-carry version of the trade solves for career risk. Most of the time, it pays a little, and occasionally it pays a huge amount.

It's probably the case that every bubble offers the economic equivalent of this, i.e. there are ways to bet that the whole thing doesn't pan out, but that if it does it's better than anyone expected. But only rarely is this something you can get in a convenient capital markets package.

This trade was truly one-of-a-kind. But the ideas behind it aren't! In The Diff, we've written about:

• The case for a concetrated portfolio.

• A breakdown of CoreWeave, where Magnetar has played a major role throughout its capital structure ($).

• How arbitrage is a great training ground for investors.

• Sometimes, structuring a trade means being able to pay in cash or GPU capacity, at many orders of magnitude ($).

• Another fun trade: the tale of Venture Global, an LNG company with artfully-crafted trades ($).

Join the discussion!

1  This one took some liberties with the question of who was the most profitable trader at Citi ($, FT), which is a problem given that the marketing hook for the book is that it's a trading memoir by the most profitable trader at Citi.

2  A fancy way to talk about equity generally is that it's the residual claimant, i.e. whatever is left over at a company after all of its formal obligations are satisfied. This is a very credit-centric way to look at the world—the people punting in quantum stocks aren't thinking in terms of residual claimants. But it's also very clarifying; a stock is a call option that wants to be a bond when it grows up, so if you're bullish and realistic, you're eventually going to look at it in bond terms.

3  I don't know why value investors in particular have this reputation, but they do. A good guess is that value investing is a somewhat moralistic exercise, where you're betting that investors are too narrative-driven to pay attention to real fundamentals, and that's why you take their money. After a while, they tend to get antsy about speculation in the stock market, or speculation somewhere else, or the deficit, or a shortage of some critical natural resource (in Burry’s case, it’s been an obsession with water). They tend to underestimate the positive feedback loop of credit expansion, where more lending leads to higher asset prices, which can be refinanced, which actually improves the perceived creditworthiness of the borrower. That dynamic describes plenty of bubbles that went on for a surprisingly long time, but it also describes the US's shift from a frontier market that was suitable only for risk capital to the world's risk-free benchmark. Sometimes, it just doesn't mean revert.