Long-Term Capital Management: The Limits of Brilliance

*A fund staffed by Nobel laureates and the finest quantitative minds available lost nearly everything in a matter of weeks. The mathematics was not the problem.*

Long-Term Capital Management was, by any measure available in 1997, the most impressive assembly of financial talent ever gathered in a single firm. Its partners included traders of formidable reputation and academics of the first rank, among them two economists who would receive the Nobel Prize for work on pricing financial instruments. Its returns in its early years were excellent, and the institutions that lent to it did so on remarkably favourable terms, reasoning that such a firm hardly required the usual scrutiny.

In the summer and autumn of 1998, the fund lost the overwhelming majority of its capital in the space of a few weeks, and its potential failure was judged sufficiently threatening to the broader financial system that the Federal Reserve Bank of New York convened its major creditors and organised a recapitalisation to permit an orderly unwinding. The episode has been studied ever since, and the central question it poses is not how such intelligent people could have been wrong, but why being right was insufficient.

The fund's core approach involved identifying securities whose prices had diverged from one another in ways the fund's analysis suggested were temporary, taking positions on both sides in the expectation that the relationship would return to its historical pattern. Individually, these positions offered small returns, because the divergences being exploited were themselves small. The mathematics identifying them was sound and the historical relationships were real. To generate substantial returns from small edges, however, the fund employed very substantial borrowing, so that a modest movement in its favour would produce a large return on its own capital.

This is the mechanism that destroyed it. Leverage does not merely amplify returns; it removes the ability to wait. An unleveraged investor whose position moves against them can hold, absorb the discomfort, and wait for their analysis to be vindicated. A heavily leveraged investor cannot, because their lenders will require additional collateral as the position deteriorates, and if they cannot provide it their positions will be closed regardless of what they believe about the eventual outcome. Leverage converts a temporary adverse movement into a permanent loss, by taking away the one thing a sound analysis requires, which is time.

The adverse movement arrived with the Russian government default of August 1998. Investors across the world sought safety, abandoning precisely the positions the fund held and rushing into the assets it had positioned against. The divergences the fund expected to narrow instead widened, and they widened simultaneously across positions its models had treated as independent.

That last point deserves emphasis, because it is the analytical failure at the heart of the episode. The fund's risk models drew on historical relationships, and in ordinary conditions those relationships held: different positions behaved differently, so that losses in some were offset by gains in others. In a period of genuine crisis, however, the relationships changed. Positions that had been unrelated became related, because they were all being abandoned by the same frightened participants for the same reason. The diversification the fund believed it possessed evaporated at exactly the moment it was needed.

It is important to note what the fund was not wrong about. Many of its positions would eventually have been vindicated, as the relationships did in time return toward their historical patterns. An investor holding them without borrowing, and with the capacity to wait, would have been proved correct. The fund had no such capacity, and so its correctness was irrelevant. This is the unforgiving lesson of the episode: an analysis that is right eventually is worth nothing to an investor who cannot survive until eventually arrives.

There is a further lesson about the relationship between models and reality. A model is built from history, and history contains only what has already happened. The events that destroy portfolios are frequently events that lie outside the historical record from which the model was constructed, and the model, by its nature, cannot warn about them. A great deal of confidence in the fund's positions rested on calculations that assigned very low probabilities to the movements that ultimately occurred. Those calculations were internally correct. They were answering a question that had been posed too narrowly.

For an ordinary investor, none of whom will ever construct anything resembling this fund's positions, the applicable lesson is nonetheless direct. Borrowing to invest removes the ability to wait, and the ability to wait is the principal advantage an ordinary investor possesses. Every argument for patience, for long horizons, for holding through difficulty, depends on not being forced to sell. Leverage is precisely the arrangement that introduces the possibility of being forced, and it does so in exchange for returns that appear attractive only when nothing goes wrong.

One further feature of the collapse deserves attention, which concerns what happens when a position becomes known. As the fund's difficulties became apparent, other participants understood both what it held and that it would be compelled to sell. This knowledge was itself damaging, because it allowed others to position themselves ahead of the selling that was known to be coming, which drove prices further against the fund and accelerated the very liquidation that was anticipated. A position that must be exited, and that is known to require exiting, will find the market unaccommodating precisely when accommodation is most needed. This is a general property of forced selling rather than a peculiarity of this episode, and it explains why the losses of a compelled seller so frequently exceed what the underlying circumstances would seem to warrant. The investor who cannot be forced to sell never encounters this problem at all, which is another way of saying that the freedom to wait is worth considerably more than it appears.

At VESTFY™ the LTCM episode is presented as the definitive demonstration that intelligence is not a substitute for survivability. The people involved were not fools; they were among the most capable participants the industry has produced, and they were undone by an arrangement that made their correctness irrelevant. Whatever else a framework does, it must first ensure that its owner is still present when their judgement is finally vindicated.