A decade before the financial crisis, we got a nice little preview of what could happen if a massively levered fixed-income trade backed mostly by short-term wholesale funding unwound. Long-Term Capital Management was a $4.7bn fund with a balance sheet that peaked at ~$125bn, and collapsed over the course of the summer of 1998, losing almost all of its' investors' money.¹ From a narrative perspective, you couldn't ask for a better story: the fund was founded by John Meriwether, who'd previously run the Salomon Brothers fixed income arbitrage business. Meriwether was the star of the opening anecdote in Liar's Poker, in which Salomon's CEO challenges him to a single hand of liar's poker (the game), for $1m. Meriwether promptly demands $10m, and the CEO folds. So you could tell a great story about hubris there. You could also tell a story about overconfident nerds (they had plenty of PhDs on staff, including Myron Scholes and Robert Merton, co-inventors of the Black-Scholes model of options pricing. And their trades and approach had the sort of manipulating-Greek-letters approach that drives more traditional investors up the wall.

So the collapse got plenty of coverage. Michael Lewis wrote a lengthy piece and Roger Lowenstein wrote a book. These both adopt a fairly moralizing tone, with Lewis focused more on the ruthlessness of other firms, which bet against LTCM, and Lowenstein incensed at all the complicated quantitative strategies they were running and the leverage they used to execute them. Lowenstein thinks they should go back to basics and read some Benjamin Graham. Graham was the dean of value investing² , but before that he was a precocious enough college student that before he finished his undergraduate degree at Columbia, the university told him he was welcome to stick around and teach math, English, or philosophy. But Graham himself also did some derivatives trading, including convertible arbitrage; he didn't turn his nose up at mathematically complex trades (he also wouldn't have been intimidated by the notation used to formalize models today, given that he could read Greek).

That narrative is broadly correct, but it's important to understand exactly what LTCM was, and the nature of the risks that killed them. There's a fiddly technical sense in which leverage reduces risk: you'll never miss a loan payment if you don't have any loans. But this is true in the same sense that never committing to anything means you'll never disappoint everyone: it's true but implies a limited existence, and also artificially flattens distinctions between useless and useful obligations. For leverage, too, the raw numbers are only relevant in light of what LTCM was levering up to do. 10:1 leverage for a trade like the futures basis trade, i.e. patiently buying relatively illiquid treasury bonds, and shorting liquid and higher-priced futures against them, is a lot safer than 2:1 leverage trading equity indices or crude oil.

And most of LTCM's trades were very much in the mold of buying one thing and shorting something else that was very similar, but not quite identical. They were trading different classes of the same stock, like Royal Dutch against Shell. They would bet on the convergence between high-interest borrowers in peripheral Europe, and hedged with short positions in German bunds. This wasn't a purely quantitative bet on correlations: it had a macro justification in that Europe was converging on a single currency and would have more consistent fiscal policy in addition to a single monetary policy so even if those bonds weren't identical, they were getting a lot more identical.

They did use plenty of fancy modeling to decide how big these bets would be, but they weren't treating those models as sacred. In fact, Victor Haghani writes in The Missing Billionaires that the firm sent a memo to its investors in October 1994 telling them that the biggest losses they'd expect were, in fact, bigger than what the models said.³ And their general approach was to avoid making any one massive bet; they were diversified across geographies, currencies, and asset classes, hedged every risk they didn't feel they were being paid appropriately to take, and tested against historically extreme downside scenarios.

But these backtests assumed the wrong constants. A levered market participant needs to negotiate leverage with its counterparties, and the counterparties have some flexibility to demand additional capital when trades start to go badly. LTCM didn't want to leak the exact trades it was making, so it would often execute one leg of a trade with one counterparty and another leg with a different counterparty. These were hedged trades, so in general when one of them was making money the other was losing, and even if the counterparty for the losing trade demanded more margin, they'd have roughly that much additional margin free from the other leg of the trade.

All of this was very sensible, even elegant, when the market was stable. And they ended up with a perverse kind of bad luck: 1997 marked the beginning of the East Asian Financial Crisis. This is the kind of disruptive market event that makes mean-reversion trades of the kind LTCM did go haywire. But they ended the year up 17%! If a good year means 40%+ performance, and a bad year means performance in the teens, it's easy to think that you've solved the market. But then, in 1998, there was yet another financial crisis, this time focused on Russia, which had defaulted on its debt. LTCM didn't have any direct exposure to Russia, but any time something weird happens, investors rush for safe, liquid assets. In other words, they sell the kinds of things LTCM owned (less-liquid bonds from less reliable credits) and buy the assets LTCM used to hedge (liquid and safe ones).

And the margin impact was uneven: they were being asked to put up much more collateral for the losses than they were given credit for from whichever legs of trades still had gains. If they'd had a bigger balance sheet, they would have been able to weather this, but they'd returned about $1.7bn in equity to their existing investors, in order to lever up the rest.

So they found themselves in a position where:

1. They were distressed, because trades were moving against them.

2. Counterparties could see this, because they were shrinking their exposure everywhere, even in trades where prices were pretty attractive.

3. Every counterparty had only part of the picture, but could trade on that part.

4. It was crash season. Presumably, at many of their counterparties the senior people were on vacation and, rather than ruin their vacation, just told underlings to cut exposure.

If you've lent someone money to make some trade (short bunds, for example), and that trade starts going bad, you can ask your counterparty for more collateral. But you can also hedge the risk that they won't be able to provide it, by taking the opposite side of the trade. And if everyone does this, the market moves in exactly the opposite direction that the models say it will.

If someone had read a few of the retrospectives on LTCM right after it happened, and then fell into a coma and woke up in the present day, there's a lot they'd find confusing, but they'd have the impression that we've learned absolutely nothing from LTCM. There are still massively levered hedge funds, and they're still doing some of the same trades LTCM did. In fact, a big one for many funds is the treasury basis trade, a classic case of a bet that goes terribly whenever markets are volatile.

Part of what these modern funds do differently is that they're much more diversified. Almost everything LTCM bet on was, fundamentally, a bet on mean-reversion. Sometimes, these trades had an expiration date—when your short futures position matures, you either roll it over or deliver the bonds you own and close it out. But some of them just bet that, at some unspecified point in the future, the world would get more normal. When a hedge fund with big exposures to equities, as well as directional commodity and macro bets adds in a few mean-reversion trades, that's defensible—they're also betting against mean-reversion elsewhere. And these funds tend to pay close attention to the liability side of their balance sheet: the more predictable portfolio managers' performance is, the more it affects outcomes if they can execute trades cheaply and efficiently, have a low cost of funding, and, critically, can keep their funding when the market goes crazy.

LTCM was ultimately bailed out, and the buyers got a good deal: they were essentially buying a very safe portfolio at 97 cents on the dollar, of which 96+ cents consisted of credit they'd already collectively extended to the firm, with the last fractional penny as equity. Summer gave way to fall, the VIX went from the 40s back to the 20s, and life went on. There was a serious risk that all the other banks running similar trades to LTCM would be dragged down with it, and that the entire financial system would be destabilized, but it turned out that the banks were able to work together, mostly, to prevent that. (Bear declined to assist in the bailout, which is one reason they were the first to fail in 2008.) So the biggest incorrect lesson of LTCM was the one we learned almost exactly a decade later: when a sufficiently big levered investor gets into trouble, they won't necessarily get rescued in time.

LTCM was an important moment in financial history—the business itself was an evolutionary dead-end, but pointed to some more viable possibilities in the future. We’ve talked about it many times:

• One piece on idea velocity notes that even though LTCM ran a diversified set of strategies, its profits were more skewed, as is the case with modern multi-strategy funds.

• We’ve speculated about what would happen if pod shops blew up.

• 1998, the year of LTCM’s collapse, was a turning point in financial history.

• You can still find some of LTCM’s marketing materials floating around.

• Don’t fear the basis trade ($).

Join the discussion!

1  Technically, the early investors did fine. Like many big funds, LTCM reached its maximum capacity and asked investors to take some of their cash back, while the principals put more money in. This is a hallowed tradition in capacity-constrained strategies, and outside investors aren't really entitled to a return unless capital is scarce. If the fund's returns are higher than the pace at which it can find new places to deploy capital, it naturally wants to return that money. But this means that investors were fully exposed during the good years, and then took money off the table before the peak.

2  Which was perhaps one of the earliest “quantitative” strategies in that it was somewhat algorithmic, but not purely quant in the statistical and systematic sense we think about it today.

3  The Missing Billionaires makes a good case for indexing and diversification, not as tools to give regular people a better retirement, but also to ensure that really rich people can have richer descendants. Haghani himself is one of those missing billionaires, someone who would be a lot richer had he not levered up to put all of his money into one investment, in this case LTCM itself.