In his latest commentary, “The Wreck of Modern Finance,” Hutchinson does a mesmerizing job of calling to task some of the “experts” (and their theories) who helped get us to where we are now.
Over the last 30 years, the capital markets have been restructured through the tenets of modern finance in its various forms, which together have gained six Nobel Prizes (Modigliani, Sharpe, Markowitz, Miller, Merton, Scholes) and might have generated a couple more (Fama and Black.) This has been enormously profitable for the financial services sector, which has doubled its share of U.S. economic output. As we are only now coming to see, it has proved pretty well disastrous for the global economy as a whole.
The most fundamental error of the modern financial edifice was its assumption of randomness in market movements, without a true understanding of the conditions necessary for randomness to hold. The laws of probability and the idea of randomness were generated by the 17th century French mathematician Blaise Pascal, who used them to help in winning card games, roulette and other activities in which tiny physical variations cause a discrete change in results. Since tiny physical variations are themselves unpredictable, their results are truly random.
This true randomness almost never holds for economic activities. Some of them are governed by complex underlying equations, impossible for mediocre mathematicians to solve, which produce pseudo-random “chaotic” behavior, in which prices or other variables appear to move randomly but are in reality mostly determinate. The interaction between economic variables is partly random (itself partly caused by inadequacies in our ability to measure economic quantities quickly) and partly determined by these kinds of complex non-linearities.
Other economic variables are also not random, and may not be governed by discoverable laws; they are simply unknown. Next year’s Gross Domestic Product, for example, is theoretically determinable from factors we could already know (plus a few random elements). But it is mostly not random; we simply do not know what it will be.
Pretending that deterministic (but chaotic) quantities or unknown quantities are random is a huge category error. If such quantities are not random, they will not obey the laws of randomness. In particular calculations that depend on the properties of randomness, such as Monte Carlo simulation, the Value at Risk methodology or the Black-Scholes options model will be quite simply wrong, often very badly wrong.
For non-random quantities, one of the most important properties of truly random quantities, the tendency of their distributions’ “tails” to disappear at around three standard deviations from the mean, will not hold. In probability, the chance of four events happening is the product of their four probabilities; in fuzzy logic, the alternative analytical system for the unknown, the “belief” of four events is the minimum of their beliefs. If each event has a probability/belief of 1 in 10, it’s the difference between 1 in 10,000 (random) and 1 in 10 (unknown).
Nassim Taleb in his best seller criticized Wall Street for being “Fooled by Randomess.” He had it precisely wrong; in reality, Wall Street and the economists and “mathematicians” (mostly not very good ones, or good ones who figured out there was a problem but stayed around for the money) have been fooled by assuming randomness where it does not exist.
That assumption makes the equations easier to solve; random quantities tend to be linear, exponential or normal, the three types of equations economists know how to deal with. Chaotic quantities generally obey power-series equations, or even nastier ones, while unknown quantities by definition don’t obey any equations at all.
Financial quantities, mostly a mixture of the chaotic and the unknown, are thus frightfully hard to model; one has some sympathy. Even Benoit Mandelbrot, a truly superior mathematician who invented fractal geometry, made devastating criticisms of others’ models in his 2004 “The misbehavior of markets,” but was unable to come up with a better alternative.
The invention of PCs, together with the intellectual “advances” of modern financial theory, from around 1980 caused an explosion of mathematical modeling on Wall Street. Models were used to price options, to value complex packages of securitized debt, to manage investments and above all, to manage risk, even being incorporated into the “Basel II” bank capital requirements.
In the “Value at Risk” risk management system, for example, the model calculates the maximum possible loss in 99% of periods covered. Even if the model worked, that would be a foolish way to manage risk for an entire bank when the periods are as short as a day or a month; 100 trading days is only five months and even 100 months is only eight years. While the 200 year life expectancies of the old merchant banks may have been excessively conservative, eight years is surely rather too short a life expectancy for banks if the market is to function soundly.
The VAR assumption, that even in the other 1% of periods the model wouldn’t be too far wrong, is completely and dangerously false. When David Vinear, chief financial officer of Goldman Sachs, said in August 2007, “We were seeing things that were 25-standard-deviation events, several days in a row,” he was condemning his risk management out of his own mouth. In a truly random system, 25-standard-deviation events would not just be rare, they would be literally impossible, of infinitesimal probability during the entire history of the universe.
Not only are price movements not random, the market is not “efficient.” It is subject to bubbles and periods of depression – indeed one of the most profitable strategies during the latter, employed by the late Sir John Templeton and others, is to buy small out-of-the-way stocks at random, since analysts stop covering the lesser names when business is bad, and their prices drift down arbitrarily far. (Conversely, that’s why many of the best investment managers, mostly invested in small stocks, did so badly in 2008, down 70% or 80% when the market overall was down only 40%; the market was de-arbitraging as the financial system fell apart.) As innumerable behavioral finance professors have demonstrated, expectations are not rational.
The Capital Asset Pricing Model also doesn’t work, as many have found to their cost. On the corporate side, it combined with the tax-deductibility of debt interest and short-term oriented compensation systems to leave banks and corporations excessively leveraged. Lehman Brothers shareholders have lost more through bankruptcy than they previously gained through leverage; business failure is in itself an extremely expensive process. Of course, bailouts here, there and everywhere have mitigated the costs of owning, say, AIG or Citigroup shares, but on the other hand, being a debt-holder in Chrysler or General Motors has proved unexpectedly expensive, as the Obama administration has reallocated resources to its union friends. In the last six months, the resource allocation process has become almost entirely political, and will remain so until the U.S. government runs out of money, which fortunately shouldn’t take too long.
On the investment side, the CAPM has led to the development of innumerable phony asset classes, whose returns were supposed to be uncorrelated to the stock market, but which were mostly notable for the hugely greater fees they provided to their sponsors. Hedge funds depend on excessive leverage and the continuance of misguided financial engineering; private equity funds depend on the existence of a thriving public market for takeout; and emerging markets funds depend on the health and liquidity of the world economy. None of them diversify risk more than marginally and all of them add huge new layers of cost. The Yale Model of investment management does not work – except for the lavishly rewarded investment managers who developed it.
Securitization appeared to be a mechanism that allowed banks to remove assets from their balance sheets, while providing relatively low-risk, liquid assets for investors. It also didn’t work. First, banks were left with most of the residual risk, so allowing them to remove the assets from their balance sheets merely encouraged excessive leverage. Second, the ratings agencies assumed randomness of outcome in, say, pools of subprime mortgage assets, an assumption that proved to be laughably wrong. When mortgage underwriting standards deteriorated, they deteriorated everywhere, so all pools ended up with their share of almost-worthless “liar loans.” Even when underwriting standards were maintained, real estate mortgage losses were not probabilistically independent, since a nationwide housing-price decline caused an ever-increasing cascade of losses, far beyond past experience. Housing and credit card loans are not probabilistically independent, because the business cycle isn’t random; in a down cycle they all go wrong at once.
The largest nirvana for mathematically-generated profits was the derivatives markets. Here the “vanilla” markets were relatively sound, with risks manageable and finite. They were, however, almost infinitely arbitrageable, so became a trading desk heaven, with far too much volume for the contracts’ real uses, endless speculative games played, and infinitesimal margins. Most important, they produced an explosion of counterparty risks so that even the dodgiest bank or broker active in the markets could not be allowed to go bust. That problem can largely be solved by President Obama’s proposed legislation forcing standard derivatives to trade over recognized exchanges, eliminating most of the counterparty risk and concentrating the rest in one place.
In order to increase profits, traders devised more and more complex derivatives types, the management of which required dangerously false assumptions about randomness. These more complex contracts up-fronted most of their profit, leading to bonus bonanzas, while leaving their risks ticking like a time-bomb through their entire duration of several years – so we may not yet have seen all the loss explosions this business will produce.
Finally, there were credit default swaps (CDS). As derivatives, these were poorly designed, because their settlement rested on a primitive auction procedure that is itself gamed by the major dealers, who use it to extract rent from the U.S. government and any companies unfortunate enough to near bankruptcy. As we have recently seen in two cases, Abitibi-Price and General Growth Properties, CDS deviate even further than most derivatives from the theoretical efficient-market modern finance ideal in that they allow debt-holders to “game” the default process itself.
In essence, CDS holders, who if they are also bondholders can vote in the bankruptcy process, act like spectators at a suicide, yelling, “Jump! Jump!” and pushing companies into default in order to reap bonanza profits from their CDS. This is particularly attractive if the CDS were issued by AIG and so effectively guaranteed by taxpayers.
CDS also act as highly efficient vehicles for short selling; their cost is so low in relation to their potential profit and their volume so large that they can provide huge incentives to unscrupulous speculators to drive viable companies over the edge – this was part of the problem at Lehman Brothers, for example.
In summary, the dangers of CDS are so out of proportion to their modest advantages as risk management tools that it seems wisest for the authorities to ban them altogether, not something I would normally recommend.
We gave poor Jeff Skilling of Enron 24 years in jail for inventing a new trading platform that turned out to be unsound. As we peer out from the wreckage that modern finance has left, one can’t help thinking that there are a number of Nobelists who more deserve such Draconian punishment.
