1. "Mark-to-market". One of the new elements in the current crisis is the requirement of banks to "mark to market". This new requirement is in response to Enron and the Japanese banking problems of the 1990's. While this requirement is well-intentioned, it has the unfortunate effect of shackling the banks together with chains, like slaves, so that when one drowns, it pulls the rest of them down, too, like dominos. When one bank is forced to sell securities at firesale/pawnshop prices, all the other banks have to mark similar securities down to the same value. If those other banks are then "under water" with respect to their required capital ratios, then those banks are forced to sell the same securities, thus driving the market down further & faster. Like the famous electrical power blackouts of the 1960's and 1970's, which took out significant portions of the United States, there are no "circuit breakers" to isolate the failure, so the crisis continues spreading until the entire system goes down. On the other hand, not eventually forcing a writedown is also bad, as the Japanese banks proved in the 1990's, because banks with truly worthless assets do need to be shut down, and any delay in their shutdown simply extends the malaise possibly for years until they are finally shut down. 2. "Collateralized Debt Obligations"="CDO's". According to last weekend's Wall Street Journal, the price of insuring Lehman debt the day before Lehman went under was $8 annually for each $100 debt outstanding. This implies that the buyers of this debt had an 8% expectation that the debt would go bad within 1 year of its purchase. The insurers (not necessarily insurance companies, but typically large financial institutions) had to put up only 10% margin on each $100 of insurance that they wrote. Thus, somone with $100 in assets could sell insurance on $1000 of Lehman debt. When Lehman went under, it triggered all of these CDO's, and apparently, nearly all of the Lehman debt _was_ insured! So, although Lehman itself was wiped out, those holders were able to transfer the losses to holders of the CDO's, who themselves may have been writing CDO's on other institutions. This is why Warren Buffet called CDO's "weapons of mass financial destruction". Late last week, CDO's on Morgan Stanley debt were selling for $16 for every $100 of debt, and the margin requirement was raised to ~20%. So writing CDO's has gotten only a little bit less risky. The interesting thing about Lehman & AIG is that Paulson shot himself (actually all of us!) in the foot (or some part of the anatomy a bit higher). The moment he took down Lehman, he guaranteed himself a much larger meltdown in other institutions -- e.g., AIG! In pre-mark-to-the-market days, this probably would not have happened with such speed & ferocity. 3. Gaussian distributions. Yes, rcs is correct, academic economists have long known about long tails. But those financial engineers in banking & commerce are still using Gaussians on a daily basis. The Black-Scholes option pricing model is a pure Gaussian model, and is calculated trillions of times a second in the supercomputers on Wall Street. This is the most classic example of "looking for your keys under the street lamp, because that is where the light is brightest". Since any other calculation is harder than Black-Scholes (hereinafter called "BS"), and since some calculation is considered better than no calculation, BS is still almost universally used. If you own stock in a publicly traded company with employee stock options outstanding, check the SEC filings; the "non-cash" stock option expenses are almost certainly computed using BS. 4. Modern portfolio theorists utilize mathematical tools like correlations to "optimize" their portfolios. If the price of one thing seems decorrelated with the price of another thing, then various financial engineering tools can be used to isolate the differences. A simple example would be choosing a single stock out of an index. Suppose you think that stock X will "do better" than its peers. You then purchase stock X and short an index fund of its peer stocks (which will have a certain amount of stock X in it). You are thus protected from movements of the index as a whole, and can focus on the single bet that "X will do better than its peers". So far, so good. But when someone constructs an "optimized" portfolio using historical data from 2004-2006 (say), one finds that the useful dimension of the space is perhaps only 15 or so, leading one to have only 15 items in the portfolio. Unfortunately, this theory doesn't take into account the possibility that the company constructing the index fund may go under, or that the market may close for a day or a week or a month (check out some exchanges like Russia, for example). And in a truly bad market, as rcs has pointed out, _everthing_ is correlated (because of massive forced sales of even "good" assets, e.g.), so your carefully constructed portfolio can be doubly screwed. Taleb teaches that with long-tailed distributions, you may need to place hundreds of bets, not just a dozen or so, in order to get a decent amount of diversification. Except that Taleb doesn't attempt to call it diversification -- he calls it being exposed to "serendipity". You may not be protected on the downside, but t he upside from some obscure investment (Youtube?) could grow exponentially to dominate your portfolio. At 02:15 PM 10/11/2008, Robert Baillie wrote:
On a related note: On Bill Moyers Journal on Oct. 10, George Soros was talking about how economists assume that things will average out and self-correct. This is, of course, false. Instead, there seems to be a tendency for self-reinforcing feedback loops to occur (in both positive and negative directions), which cause exponentially fast changes to occur. Unforunately, that's often too fast for society or the political system to react.