"
Thus the valuation of a complex financial engineered product (i) may not be generally agreed among market participants, (ii) may quite simply be wrong (iii) may be proved hopelessly flawed by new discoveries about the underlying asset class or (iv) may be affected by distorted incentives so that the owners of the product, the banks concerned, receive different rewards from the agents controlling the product, the executives. These problems may lie dormant for a decade or more and then manifest themselves sharply in periods of market turbulence, causing confidence among market participants to vanish.
Finally the risk management models used by institutions to control the risks of their financially engineered holdings are themselves hopelessly flawed, particularly the Value-At-Risk system. Under VAR, the risk of an asset holding is calculated as the maximum fluctuation in the value of that holding in 99% of cases. The gross assumption is then made that price movements are normally distributed, so the risks in the other 1% of cases can be assumed to be only modestly greater than the stated VAR. As Goldman Sachs showed, in announcing a “25 standard deviation” event that should under VAR assumptions happen only once in the life of the universe, this is just plain wrong. It again rests on the flawed underlying postulate that market events are random, which has repeatedly been shown to be in many cases false.
...
The profitability of financial engineering to its practitioners is unquestionable. Its profitability to the institutions that employ those practitioners may have been almost equally solid in the past, but could be undermined in the future by a period of market turbulence that produces gigantic write-offs – a house like Goldman Sachs, with “Level 3” assets, the most illiquid, of twice its capital could in principle suffer losses that wiped out all its financial engineering profits of the last quarter century.
Financial engineering’s benefit to the global economy is questionable at best and the increases it has produced in the financial services sector’s share of global output may have been mere successful rent seeking. In the long run, less opulent compensation for financial engineers, more aggressive audit and supervision policies for financial institutions’ engineered assets and a healthy cynicism about financial engineering in general may put this genie at least half way back into its bottle. That is likely to prove a positive development."
Why Financial Engineering Doesn’t WorkBy Martin Hutchinson
November 5, 2007
SourceMartin Hutchinson is the author of "Great Conservatives" (Academica Press, 2005) -- details can be found on the Web site www.greatconservatives.com
The last quarter century has seen the explosion of a profession, financial engineering, that has provided innumerable lucrative opportunities for otherwise indigent mathematicians -- it was thus welcomed by a former mathematician like myself! Nevertheless the turbulence in the bond markets in the last couple of months, at a time when the world economy’s prospects seemed set fair, have exposed a guilty secret of the financial engineering profession: its methods don’t work.
The first exposure of financial engineering’s failings came oddly enough in the area that had seemed most solid, that of liquidity. Theoretically, if financial engineers design ever fancier artificial securities and derivatives, but everybody uses the same mathematical models to value them, there is no reason why an active market should not operate in the securities, at whatever price the models direct.
In practice this only works in calm markets. In times of market turbulence, when doubts arise about either the mathematical models themselves or the underlying assets from which value is derived, the true value of these artificial assets becomes thoroughly unclear. Buyers assess their value at the lowest possible level, and refuse to increase their exposure to a sector that suddenly appears dangerous, while sellers attempt to get out of the business altogether.
That explains the sudden drying up in support for asset backed commercial paper in August. It also explains the precipitate drop in the prices of ABX credit default swaps on subprime mortgage backed securities in October. Overall the market in such swaps declined by no less than 25 percent, while AA rated credit default swaps, supposedly among the finest credits available, were trading at less than 50% of principal amount by the end of the month. Either the rating agencies had gone horribly wrong in assessing the default risk of those AA credit default swaps, or the underlying pool of subprime mortgages was so rotten that more than half of a $1.5 trillion asset class would vaporize. The real problem for the market was: nobody knew which.
The problem was further compounded by Wall Street’s accounting methodology, which allowed assets to be valued based on the theoretical prices produced by the mathematical model. If as in this case reality had made already illiquid assets impossible to value, the mathematical model’s prices could become wildly out of line with reality, which was itself unknowable. Naturally, Wall Street institutions did not wish to take writedowns on the assets, so were unwilling to undertake transactions at prices which might call into question the valuation methodology of their portfolio. The effective market for the assets thus settled to something like 20 bid, 90 offered, killing the price discovery process and turning them into toxic waste on their owners’ balance sheets.
Banks and investment banks have adopted a number of approaches to the problem of their balance sheets’ subprime mortgage-related assets. Merrill Lynch wrote the assets down by $4.9 billion at the end of September, but then found itself obliged to write them down by a further $3.5 billion when publishing their third quarter figures in late October, thus bringing the departure of chief executive Stanley O’Neal. Given the ABX price drop in October, a further writedown may now be needed. Goldman Sachs maintained a posture of insouciance, claiming that the brilliance of their hedging had prevented any significant losses at all (but were the hedges adequately liquid, or was their increased value simply the result of aggressive revaluation through mathematical models?) Sachsen LB of Germany threw up its hands at the impossibility of funding its subprime mortgage portfolio and subsided into the arms of a larger bank, Landesbank Baden-Wurttemburg.
Only Nomura Securities took a properly decisive approach, selling its entire $2.4 billion portfolio of mortgage based assets (thus worsening the problem for everybody else) firing the people involved and taking a $700 million write-off, about 28% of its mortgage assets, and slightly more than 100% of its subprime portfolio. Writing off 105% of one’s holdings of a financial engineering product may be regarded as aggressive, but its cold realism is probably healthy in the long run.
The second problem with financial engineering products that the whizzes involved have failed to solve is their valuation. Simple derivatives such as interest rate swaps in currencies with liquid government bond markets and forward Treasury futures can be valued by mathematical techniques that are simple and fairly robust; thus market participants can agree on the underlying value of these assets even when markets are turbulent.
The problem becomes much more difficult when a financially engineered product involves an embedded option, or like asset backed commercial paper is of markedly different liquidity from its underlying asset. In the case of liquidity differential there is no generally agreed way to value liquidity, hence no means of ensuring that mathematical models account properly for the difference in liquidity between short term and long term securities. Creating artificial liquidity through use of structured investment vehicles may produce an apparent stream of income to the arranger from the differential between short term and long term yields, at the risk of a huge liquidity crisis in the event that short term paper can no longer be sold at a reasonable price.
The standard option valuation models do not work at all well for out-of-the-money options, because they assume the randomness of future events that are in reality not random but unknown. The phenomenon of volatility “smile” in option pricing, whereby the implied volatility of out-of-the-money options is considerably higher than that of at-the-money options, is a sign that the underlying theory, which postulates constant volatility over the full range of strike prices, is hopelessly flawed.
The valuation problem is worsened for financially engineered products with multiple embedded options. In these cases, some of the options are at-the-money, some in-the-money and some out-of-the-money. Even when the majority of market participants are using valuation models that produce similar answers, those models may bear little relationships to true market value. This problem is worsened when the characteristics of an asset class upon which its derivatives’ valuation is based come seriously into question.
In the subprime mortgage case, for example, it had been assumed that mortgage defaults were essentially independent of each other, enabling valuers to use “laws of large numbers” to “prove” the probability of a default of more than 20% of principal was very small. That allowed securities based on that senior slice of the assets to be rated AAA. In reality, defaults on subprime mortgages are not independent events. A mortgage bubble such as that of 2004-06 causes a simultaneous slackening of underwriting standards, with even minimal control procedures being abandoned throughout the entire asset class, while a nationwide house price decline or interest rate rise causes the entire class of subprime mortgages to get into simultaneous difficulty.
Finally, some trading strategies are particularly attractive to market participants because they pull income up-front, enabling participants to recognize larger profits (and presumably receive larger bonuses) in the current year while deferring losses into future periods when they may have left the group.
Thus the valuation of a complex financial engineered product (i) may not be generally agreed among market participants, (ii) may quite simply be wrong (iii) may be proved hopelessly flawed by new discoveries about the underlying asset class or (iv) may be affected by distorted incentives so that the owners of the product, the banks concerned, receive different rewards from the agents controlling the product, the executives. These problems may lie dormant for a decade or more and then manifest themselves sharply in periods of market turbulence, causing confidence among market participants to vanish.
Finally the risk management models used by institutions to control the risks of their financially engineered holdings are themselves hopelessly flawed, particularly the Value-At-Risk system. Under VAR, the risk of an asset holding is calculated as the maximum fluctuation in the value of that holding in 99% of cases. The gross assumption is then made that price movements are normally distributed, so the risks in the other 1% of cases can be assumed to be only modestly greater than the stated VAR. As Goldman Sachs showed, in announcing a “25 standard deviation” event that should under VAR assumptions happen only once in the life of the universe, this is just plain wrong. It again rests on the flawed underlying postulate that market events are random, which has repeatedly been shown to be in many cases false.
VAR’s underestimation of risk is particularly severe for financially engineered products that have large numbers of embedded options, or that depend on an asset class such as subprime mortgages with extreme risk characteristics. The problem is exacerbated by the valuation uncertainty of such products, and by their tendency to become completely illiquid in times of market turbulence. Thus two balance sheets with an equal VAR may have a very different level of risk; the institution that has been more aggressive in its financial engineering activity is likely to be much riskier than the other. Even if an aggressive trading house and a conservative new-product-averse commercial bank claim similar levels of VAR in their portfolios, the trading house’s true risk is likely to be much higher, because it will have a higher concentration of aggressively engineered assets with numerous embedded options, flaky underlying assets and severe turbulent-market liquidity risk.
The profitability of financial engineering to its practitioners is unquestionable. Its profitability to the institutions that employ those practitioners may have been almost equally solid in the past, but could be undermined in the future by a period of market turbulence that produces gigantic write-offs – a house like Goldman Sachs, with “Level 3” assets, the most illiquid, of twice its capital could in principle suffer losses that wiped out all its financial engineering profits of the last quarter century.
Financial engineering’s benefit to the global economy is questionable at best and the increases it has produced in the financial services sector’s share of global output may have been mere successful rent seeking. In the long run, less opulent compensation for financial engineers, more aggressive audit and supervision policies for financial institutions’ engineered assets and a healthy cynicism about financial engineering in general may put this genie at least half way back into its bottle. That is likely to prove a positive development.