Sunday, September 05, 2010

Book review - When Genius Failed

This book on the LTCM debacle is by no means either new or contemporary. I just happened to dig it out from Landmark one evening and finished reading it today.

LTCM i.e. Long Term Capital Management, was a hedge fund promoted by a clutch of bond traders on Wall Street along with a few academicians (including Scholes and Merton of Black-Scholes-Merton fame). At its peak it had equity capital of more than $ 4 bn and assets of more than $ 100 bn.

The book covers the rise and fall of LTCM in great detail - in a very human rather than analytical way. It is written for lay persons and that shows. A good read for lay persons for sure, but also for the practitioners of modern finance, all the more so in emerging economies like India's. It is not implausible to expect an LTCM like debacle in India within next 5 years. As the complexity of the financial system grows, crises of LTCM variety are more and more likely to occur.

So what exactly happened? LTCM started out as purely a bond arbitrage portfolio with $ 1.5 bn in capital. It borrowed about 25 times that and invested in bonds of various types. It was not a directional bet on the bond prices though. It dealt in what is called convergence trades i.e. narrowing of spreads between two bonds. For whatever reasons if the spreads between two bonds are out of whac with the usual levels, an opportunity to make relatively low risk return arises. One needs to short the costlier bond and long the cheaper one. Over time, as the spreads do narrow, one squares off both the positions and makes a tidy profit. This is market neutral trade since absolute price levels of either of the two positions is not relevant for the profit and loss of the trade. Only the spreads matter.

LTCM did good business in first four years of its operation. However, its success prompted many others to enter the bond arbitrage business. This put the firm in a quandary. It was getting more investors but fewer opportunities. The firm reacted by diversifying into newer areas of convergence trade - including merger arbitrage, volatility spreads and so on. This however was a much more tricky turf than bond arbitrage. Owing to a blind faith in the validity of their lognormal price models and a financial storm precipitated by the Russian bond default of the 1998, LTCM found that the spreads in most of its trades were diverging rather than converging. Owing to its excessive leverage, this proved to be far too much for it to handle. The fund had to be rescued by a consortium of investment banks.

The efficient market hypothesis has turned out to be more a neat modeling tool rather than reliable predictive framework. The book repeatedly refers to fat tails and the ridiculous implicit expectation in normal distribution models that events of 1998 should occur no more than once in the lifetime of the universe! In reality, such events happen once in 25-50 years; and are occuring more frequently in recent decades.

An interesting meta-inference i could draw was as follows. The accuracy of the efficient market hypothesis itself is a market dependent variable. Think of it as the volatility of the volatility. During normal times, the normal distribution (no pun intended) holds much more accurately than during turbulent times. I recall my lessons as an aerospace engineer. In fluid mechanics we always specified whether our models were dealing with laminar flow or turbulent flow (when you open a tap just a little, its water stream is laminar, if you open it fully and the water pressure is good, it becomes turbulent). The models, one would guess, differed significantly. Something similar should apply to the financial world as well. The laminar models are already in place. The turbulent ones need to be developed. (Jump diffusion for option pricing is an interesting start.)

No comments: