Tue Dec 25th, 2007 at 11:37:10 AM EST
Over the past six months or so we have had numerous discussions of the unfolding subprime crisis and of the credit crunch and liquidity freeze that have gripped the financial markets since the summer. It is not unlikely that in the coming year we will see some banks fail, the bottom fall out of the housing market in several countries, a recession and possibly even a depression if things get really out of hand. And out of hand they might well get. The media and the blogs are awash with commentary on the failure of the US Federal reserve to make a dent in the credit/liquidity crisis even with interest-rate cuts which would be quite bold on their own but seem outright desperate when there is a widely acknowledge risk of inflation. Only even bolder (the FT called them shocking) money injections by the European Central Bank have been able to ease the interbank markets, but only until the New Year.
In these discussions we have touched on the problem of over-reliance on mathematical models, including extrapolating them to regimes where the parameters haven't been calibrated (using a model in "uncharted waters") and valuing large chunks of banks' balance sheets by marking to model, that is, with no reference to a market price because in some cases there just isn't a market for the assets. We have also learned that the US has accounting rules that require companies to classify their assets into "levels", of which "level 3" is "valued using management's best estimate" and that some of the most powerful banks have more "level 3" assets than net equity (in other words, if the level 3 assets turn out to be nearly worthless, the banks would be technically bankrupt).
For this reason it is nothing less than infuriating to look at some of the textbooks that those of us at the bottom of the quantitative finance food chain have to read in order to make it through an interview and find that clearly worded warnings about all this are right there in black and white on the pages of these books. It makes me wonder whether people, including senior people, just read the books for the formulas and eschew the qualitative insight.
Follow me below the fold for some quotations that show that it needn't have been this way.
The first of the books I want to quote from is John C. Hull's Options, Futures and Other Derivatives. The last (30th) chapter of the book is called Derivatives mishaps and what we can learn from them. Clearly what can be learned has not been learned because the book says the following:
LESSONS FOR FINANCIAL INSTITUTIONS
Do not blindly trust models
Some of the large losses in Table 30.1 [Big losses by financial institutions] arose because of the models and computer systems being used. ...
If large profits are being made by following relatively simple trading strategies, there is a good chance that the models underlying the calculations of profits are wrong. Similarly, if a financial institution appears to be particularly competitive on its quotes for a particular type of deal, there is a good chance that it is using a different model from other market participants, and it should analyze what is going on carefully. ...
Be conservative in recognizing inception profits
When a financial institution sells a highly exotic instrument to a nonfinancial corporation, the valuation can be highly dependent on the underlying model. ... In these circumstances, a phrase used to describe the daily marking to market of the deal is marking to model. This is because there are no market prices for similar deals that can be used as a benchmark.
Suppose that a financial institution manages to sell an instrument to a client for $10M more than it is worth—or at least $10M more than the model says it is worth. The $10M is known as inception profit. When should it be recognized? ... Some [investment banks] recognize the $10M immediately, whereas others are much more conservative and recognize it slowly over the life of the deal.
Recognizing inception profits immediately is very dangerous. It encourages traders to use aggressive models, take their bonuses, and leave before themodel and the value of the deal come under close scrutiny. ...
Do not sell clients inappropriate products
It is tempting to sell corporate clients inappropriate products, particularly when they appear to have an appetite for the underlying risks. But this is shortsighted. ... The [particularly risky] products worked well for [Bankers Trust]'s clients in 1992 and 1993, but blew up in 1994 ... The bad publicity hurt BT greatly. The years it had spent building up trust among corporate clients and developing an enviable reputation for innovation in derivatives were largely lost as a result of the activities of a few overly aggressive salesmen. ...
Do not ignore liquidity risk
Financial engineers usually base the pricing of exotic instruments and instruments that trade infrequently on the prices of actively traded instruments. ...
This practice is not unreasonable. However, it is dangerous to assume that less-actively traded instruments can always be traded at close to their theoretical price. When financial markets experience a shock of one sort or another there is often a "flight to quality". Liquidity becomes very important to investors, and illiquid instruments often sell at a big discount to their theoretical values. Trading strategies that assume large volumes of relatively illiquid instruments can be sold at short notice at close to their theoretical values are dangerous.
Beware when everyone is following the same trading strategies
It sometimes happens that many market participants are following essentially the same trading strategy. This creates a dangerous environment where there are liable to be big market moves, unstable markets, and large losses for the market participants.
Hmm... "contagion" caused by one distressed player "unwinding" positions which many other players are holding as well? Liquidity squeezes? Problems caused by Mark-to-model in its many forms? Problems caused by inappropriate accounting recognition of profits and losses? Blind trust in models, including using them outside the domain where they have been calibrated? We've seen it all in the last
, and it is all there in this chapter of the book everyone is supposed to have read! (Admittedly, this chapter was added to the 5th edition in 2003, but still...)
The second book I want to quote from is Nassim Taleb's 1997 Dynamic Hedging: Managing Vanilla and Exotic Options which is nowadays marketed as "Taleb on Risk". This is a very technical practical book but it is also full of qualitative insight of a general nature. And it has been in print (and sold very well) for the past 10 years. So, unlike that last chapter of Hull which was only added in 2003, veteran risk managers cannot claim that they were not aware of the stuff in Taleb's book. And yet...
Here's, for starters, Taleb's take on inception profit. Compare the practitioner's approach with the professor's above.
Market Making and the Illusion of Profitability
Market making in complex products can show immediate rewards and future thorns. Mostly profits are shown after the trade because the mark-to-model process generally implemented derives security prices off a mid-market. Traders usually find a fair value and mark the derivative up, then recognize most or all of the difference between fair value and the selling price as a profit. This method does not allow for the future costs incurred in the management of an option book. The more complex the security, the more subsequent dynamic hedging will be required, and that in more than one market ... Heavy slippage will be incurred when the position attains some large size. All this would eat up the initial profit, but would often be concealed from the trader who receives a steady flow of such profit-booking trades. It would be difficult for the trader to espouse a more aggressive stance and reduce transaction costs by observing the different behaviour of the multifaceted components. Market makers are crippled by the size of their positions in relation to their appetite for risk.
it's all there, isn't it? The dangers of mark-to-model, the hidden long-term costs of complex financial products very profitable in the short term (this, again, brings up issues of accounting practice). The fact that it is hard to resist getting into too large positions when the market demands it and the initial profit is large... And this is in the introductory, least technical part of the book.
We then come to liquidity risk. Taleb devotes an entire chapter to liquidity risk, to the way that liquidity holes take market behaviour outside the scope of standard models of market behaviour, and to the way liquidity can be manipulated by traders.
If one were to summarize what trading (as opposed to investing) is about, the best answer would be adequate management (and understanding) of liquidity. Liquidity is the source of everything related to markets
So far, so good. Now remember what Hull says: do not underestimate liquidity risk.
- A Liquidity hole or a black hole is a temporary event in the market that suspends the regular mechanics of equilibrium attainment. It is an informational glitch in the mechanism of free markets, one that can cause considerable damage to firms. In practice, it can be seen when lower prices bring accelerated supply and higher prices accelerated demand.
I guess the point is that this is not a rare event
. Markets are punctuated by these shocks on a regular basis. Only sometimes the size of the jolt is large enough to make the news.
Liquidity holes are attributable to the way information initially affects a market. Typically, liquidity holes occur when operators are aware of a major piece of information (an event, or a size order in the market) but cannot gauge its size and possible impact. Information causes anxiety, conspiracy theory, and price conflicts. Most often, operators need to interpret information. Markets are supposed to move to the extent of the difference between news and initial expectations but the latter are usually unknown and difficult to estimate properly.
No theory survives first contact with experiment?
Liquidity and Risk Management
It cannot be stressed enough that liquidity is the most serious risk management problem. A substantial part of unforeseen losses is due either to market jumps caused by illiquidity or to liquidation costs that substantially move the market against one's position. Liquidation costs tend to be usually underestimated since operators usually "fade" when someone is forced into a market action
Some of you may remember that the credit crunch started in the summer with a lot of "turbulence" in the markets caused by hedge funds exposed to subprime meeting margin calls by selling off positions in liquid assets which moved everyone else's portfolios (see above "beware of everyone following the same strategy") in the wrong direction triggering more margin calls and further liquidations. When Taleb talks about "operators fading" he means this:
When a stop-loss is executed ... operators freeze as they become aware of the order without further information as to its magnitude Many suspend their trading temporarily, thus causing the market to gyrate even further.
This is because withdrawing from the market lowers the liquidity and the lower the liquidity the more sensitive the market price is to individual orders, that is, noise.
Taleb then gives an example of a trader manipulating the market by using his knowledge of a stop-loss order by a client, and deliberately causing a liquidity hole in the process, as if it were the most normal thing in the world. To me, this example give a whole new meaning to Chris Cook's distinction between systematic and systemic market manipulation. [See, for instance: Acceptable Market Manipulation by ChrisCook on November 9th, 2006]
But the scary bit comes next, when Taleb argues that exotic options just make matters worse!
Much of the recent [1993-6?] market volatility was caused by the triggering of barrier options. As these are starting to abound, so are unexplainable gaps in the markets beginning to appear. In the previous example, we saw that a large stop was a free option. However, a stop can be canceled, or given to a trader too late for him to be able to utilize it properly. A trigger option will offer none of these mitigating features: it will be there offered to the market maker far in advance and by its nature is noncancellable.
I call this a Lazy Quote Diary because I don't have much of my own to say except that the more I learn about this stuff, the more shocked I am that anyone could be shocked at what has been happened in the financial markets over the past six months, least of all people who write for a living on the pages of the financial press.