I’ve just finished reading a book I probably should have read years ago. It’s When Genius Failed, by Roger Lowenstein. It was published in 2000, so I’m well behind the times… Anyway, it’s the story behind the rise and fall of Long-Term Capital Management, and it’s a fascinating study of how the world of finance managed to follow a very dubious detour under the influence of some highly-respected economists and mathematicians - including Nobel Prizewinners.
Those of my generation and older will remember how a bunch of US academics - Merton Miller, Robert Merton, Fischer Black and Myron Scholes amongst them - began harnessing the growing power of computers to model financial markets in the late 1970s. Out of that came the Black-Scholes option pricing model and the concept of Value at Risk as a way of measuring a company’s risk position. Two of the central planks of their philosophy were the random walk theory and the efficient markets hypothesis; more of those in a moment.
LTCM morphed out of the bond arbitrage group at Salomon Brothers in the early nineties. That business group - headed by a man named John Meriwether - had been remarkably successful, generating a significant part of the bank/brokerage’s profits for a large part of the 1980s. Meriwether had embraced the scientific approach of those rockstar academics - and indeed employed some of them and their cohorts - and ran the bond arbitrage business relying on computer modelling rather than the traditional traders’ gut feeling about markets. Unfortunately, one of Meriwether’s underlings played silly games with US government bond auctions, and Meriwether - who may or may not have been aware of it - had to carry the can and left Salomon.
But that just freed him up to found LTCM, employing many of his former Salomon colleagues and adding real stardust by persuading Robert Merton and Myron Scholes (as well as a former Fed deputy director) to take partnerships in the fund. And they went off like a rocket; they raised more finance than any fund had ever done before on its opening, and they started putting in extraordinarily impressive returns to their investors. That lasted for about four years, and then it all went horribly wrong and imploded in a matter of months in 1998, necessitating the Fed banging together the heads of LTCM’s banks to cobble together a rescue of sorts. The fear was that the fund had grown so big, and so enormous were its exposures to most of Wall Street that it could endanger the whole financial system.
So that was the basic story. What’s interesting is why this seemingly successful fund, run by heavy-duty intellects, went wrong, because the partners had been so sure that their models could accurately predict financial markets and that therefore - although they would sometimes suffer losses - they would not be exposed in such a catastrophic way as to destroy them.
To understand, we need to go back to those two bases - random walk theory and the efficient markets hypothesis. Random walk is basically coin tossing. If you toss a coin, there are only two possible outcomes, and however many times you toss it, each time, the chances are 50/50, heads or tails. The coin has no memory, and if you toss three tails, the probability on the next throw is still 50/50. We can expand this a bit, and look at rolling two six-sided dice. The most likely outcome of this is that you will throw a seven (because there are more ways to make seven than any other total) and it’s easy enough to calculate all of the probabilities. If we then plot all this on a graph it will result in that well-known shape, the bell curve of the normal distribution. That clusters the most likely outcomes around the middle and leaves the outliers under the (flat) tails on either side.
The efficient markets hypothesis is the theory that asset prices always reflect the sum of the information available at any given time. Therefore mis-pricing is theoretically impossible. There are wrinkles involving strong, semi-strong and weak variants, largely to do with whether ‘insider’ information is also impacting, but we don’t need to go into that here. (Just as an aside, these concepts took a strong hold; I took my MBA in the early nineties, and the finance courses were almost exclusively based on this stuff. The lecturers included some pretty high-powered asset managers, and questioning the validity of mathematical modelling in finance was not the way to score highly at the time.)
Anyway, taken together, these models represent the ‘rational market’, and LTCM, with its battery of major financial economists, was in a way the poster child for this form of money management. When markets do what they are supposed to do, it all works well. But when they don’t, the models prove sadly inadequate. So LTCM began making massive returns (added to by high leverage, which we’ll look at in a moment), but when things began to become unpredictable - principally caused by the Russian and Asian debt crises of the late 1990s - they found that they were utterly at the mercy of markets which did not behave as the models said they would.
I think there is one over-riding reason for this. The coin and the dice have no memory - that’s clear - so they will behave in a random way and over sufficient tosses or rolls will fit the normal distribution. However, to jump from that to an assumption that markets also have no memory is incorrect. Traders will not behave in a purely rational way - fear, greed, panic: human emotions are an integral part of how they work, and they cause irrational movements of market prices. That has significant implications. Think of the Black/Scholes option pricing model - the apogee of mathematics in markets - and its dynamic hedging. And then remember the times when it is impossible to keep up with a (most often) falling market that is running away from the delta you need to sell as liquidity dries up. As an example of this, I can think of the day the Sumitomo scam became apparent. Like many others, as the price had been rising over previous months and years, we had been selling put options to copper producers eager to take some benefit from the price strength. As the price rose, so the delta of those options decreased, creating space to sell more. But when the dam broke, the price dropped and the delta position started growing rapidly, and there wasn’t the liquidity in the market to absorb the selling. Anybody who had been over-leveraged would have been unable to survive the mark-to-market exposure created, precisely because the model fails to understand that markets are not efficient and rational. That’s the effect of human emotion, and that’s what is behind the comment Keynes made (which the LTCM boys would have done well to heed ) - “The market can stay irrational longer than you can stay solvent”.
Which brings us neatly to the last point I’m going to make - the pernicious effect of leverage. LTCM raised a lot of money, for sure, but in order to make real money out of wafer thin bond arbitrage spreads, they also borrowed a lot. Really a lot. They were leveraged at times up to 50 times (in the billions, and not including the derivative positions…), and 28 to 30 seems to have been the norm. As to why they were able to persuade the banks to loan them so much, and with so little collateral being demanded, I’m not really going to comment, but good old fear and greed came in to it. The result, though, was that when markets span away from them, they had massive asset positions supported by a disturbingly thin slice of capital. Quite aside from the fallible modelling, that demonstration of the danger of being over-extended by excess borrowing should be a wake-up call for anybody. (In fact, it would be good if politicians - are you listening, Corbyn and McDonnell? - were to look at the dangers of cavalier borrowing…)
There’s a lot more to say about LTCM, and I’ve only scratched the surface of some of the issues. But I strongly recommend the book - it’s a good lesson in arrogance and hubris.
‘When Genius Failed’, by Rupert Lowenstein, is published by Fourth Estate, and also available on Amazon Kindle.